update readme, add retinafaceAntiCov
This commit is contained in:
wang-xinyu
parent
b95b0a4e26
commit
5273421dda
@ -6,7 +6,7 @@ So why don't we just skip all parsers? We just use TensorRT network definition A
|
||||
|
||||
I wrote this project to get familiar with tensorrt API, and also to share and learn from the community.
|
||||
|
||||
All the models are implemented in pytorch first, and export a weights file xxx.wts, and then use tensorrt to load weights, define network and do inference. Some pytorch implementations can be found in my repo [Pytorchx](https://github.com/wang-xinyu/pytorchx), the remaining are from polular open-source pytorch implementations.
|
||||
All the models are implemented in pytorch or mxnet first, and export a weights file xxx.wts, and then use tensorrt to load weights, define network and do inference. Some pytorch implementations can be found in my repo [Pytorchx](https://github.com/wang-xinyu/pytorchx), the remaining are from polular open-source implementations.
|
||||
|
||||
## News
|
||||
|
||||
@ -50,6 +50,7 @@ Following models are implemented.
|
||||
|[yolov4](./yolov4)| CSPDarknet53, weights from [AlexeyAB/darknet](https://github.com/AlexeyAB/darknet#pre-trained-models), pytorch implementation from [ultralytics/yolov3](https://github.com/ultralytics/yolov3) |
|
||||
|[retinaface](./retinaface)| resnet-50, weights from [biubug6/Pytorch_Retinaface](https://github.com/biubug6/Pytorch_Retinaface) |
|
||||
|[arcface](./arcface)| LResNet50E-IR, weights from [deepinsight/insightface](https://github.com/deepinsight/insightface) |
|
||||
|[retinafaceAntiCov](./retinafaceAntiCov)| mobilenet0.25, weights from [deepinsight/insightface](https://github.com/deepinsight/insightface), retinaface anti-COVID-19, detect face and mask attribute |
|
||||
|
||||
## Tricky Operations
|
||||
|
||||
|
||||
41
retinafaceAntiCov/CMakeLists.txt
Normal file
41
retinafaceAntiCov/CMakeLists.txt
Normal file
@ -0,0 +1,41 @@
|
||||
cmake_minimum_required(VERSION 2.6)
|
||||
|
||||
project(retinafaceAntiCov)
|
||||
|
||||
add_definitions(-std=c++11)
|
||||
|
||||
option(CUDA_USE_STATIC_CUDA_RUNTIME OFF)
|
||||
set(CMAKE_CXX_STANDARD 11)
|
||||
set(CMAKE_BUILD_TYPE Debug)
|
||||
|
||||
find_package(CUDA REQUIRED)
|
||||
|
||||
set(CUDA_NVCC_PLAGS ${CUDA_NVCC_PLAGS};-std=c++11;-g;-G;-gencode;arch=compute_30;code=sm_30)
|
||||
|
||||
include_directories(${PROJECT_SOURCE_DIR}/include)
|
||||
if (CMAKE_SYSTEM_PROCESSOR MATCHES "aarch64")
|
||||
message("embed_platform on")
|
||||
include_directories(/usr/local/cuda/targets/aarch64-linux/include)
|
||||
link_directories(/usr/local/cuda/targets/aarch64-linux/lib)
|
||||
else()
|
||||
message("embed_platform off")
|
||||
include_directories(/usr/local/cuda/include)
|
||||
link_directories(/usr/local/cuda/lib64)
|
||||
endif()
|
||||
|
||||
|
||||
set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -std=c++11 -Wall -Ofast -Wfatal-errors -D_MWAITXINTRIN_H_INCLUDED")
|
||||
|
||||
cuda_add_library(myplugins SHARED ${PROJECT_SOURCE_DIR}/decode.cu)
|
||||
|
||||
find_package(OpenCV)
|
||||
include_directories(OpenCV_INCLUDE_DIRS)
|
||||
|
||||
add_executable(retinafaceAntiCov ${PROJECT_SOURCE_DIR}/retinafaceAntiCov.cpp)
|
||||
target_link_libraries(retinafaceAntiCov nvinfer)
|
||||
target_link_libraries(retinafaceAntiCov cudart)
|
||||
target_link_libraries(retinafaceAntiCov myplugins)
|
||||
target_link_libraries(retinafaceAntiCov ${OpenCV_LIBS})
|
||||
|
||||
add_definitions(-O2 -pthread)
|
||||
|
||||
45
retinafaceAntiCov/README.md
Normal file
45
retinafaceAntiCov/README.md
Normal file
@ -0,0 +1,45 @@
|
||||
# RetinaFaceAntiCov
|
||||
|
||||
The mxnet implementation is [deepinsight/insightface/RetinaFaceAntiCov](https://github.com/deepinsight/insightface/tree/master/RetinaFaceAntiCov).
|
||||
|
||||
## Run
|
||||
|
||||
```
|
||||
1. generate retinafaceAntiCov.wts from mxnet implementation.
|
||||
|
||||
git clone https://github.com/deepinsight/insightface.git
|
||||
cd insightface/RetinaFaceAntiCov
|
||||
// download its weights 'cov2.zip', put it into insightface/RetinaFaceAntiCov, and unzip it
|
||||
// put tensorrtx/retinafaceAntiCov/gen_wts.py into insightface/RetinaFaceAntiCov
|
||||
python gen_wts.py
|
||||
// a file 'retinafaceAntiCov.wts' will be generated.
|
||||
|
||||
2. put retinafaceAntiCov.wts into tensorrtx/retinafaceAntiCov, build and run
|
||||
|
||||
git clone https://github.com/wang-xinyu/tensorrtx.git
|
||||
cd tensorrtx/retinafaceAntiCov
|
||||
// put retinafaceAntiCov.wts here
|
||||
mkdir build
|
||||
cd build
|
||||
cmake ..
|
||||
make
|
||||
sudo ./retinafaceAntiCov -s // build and serialize model to file i.e. 'retinafaceAntiCov.engine'
|
||||
wget http://www.kaixian.tv/gd/d/file/201611/07/23efff3a26e2385620e719378c654fb1.jpg -O test.jpg
|
||||
sudo ./retinafaceAntiCov -d // deserialize model file and run inference.
|
||||
|
||||
3. check the images generated, as follows. out.jpg
|
||||
```
|
||||
|
||||
<p align="center">
|
||||
<img src="https://user-images.githubusercontent.com/15235574/84776553-069c5f80-b013-11ea-893c-70a138b843d6.jpg">
|
||||
</p>
|
||||
|
||||
## Config
|
||||
|
||||
- Input shape `INPUT_H`, `INPUT_W` defined in `decode.h`
|
||||
- FP16/FP32 can be selected by the macro `USE_FP16` in `retinafaceAntiCov.cpp`
|
||||
- GPU id can be selected by the macro `DEVICE` in `retinafaceAntiCov.cpp`
|
||||
|
||||
## More Information
|
||||
|
||||
See the readme in [home page.](https://github.com/wang-xinyu/tensorrtx)
|
||||
227
retinafaceAntiCov/decode.cu
Normal file
227
retinafaceAntiCov/decode.cu
Normal file
@ -0,0 +1,227 @@
|
||||
#include "decode.h"
|
||||
#include "stdio.h"
|
||||
|
||||
namespace nvinfer1
|
||||
{
|
||||
DecodePlugin::DecodePlugin()
|
||||
{
|
||||
}
|
||||
|
||||
DecodePlugin::~DecodePlugin()
|
||||
{
|
||||
}
|
||||
|
||||
// create the plugin at runtime from a byte stream
|
||||
DecodePlugin::DecodePlugin(const void* data, size_t length)
|
||||
{
|
||||
}
|
||||
|
||||
void DecodePlugin::serialize(void* buffer) const
|
||||
{
|
||||
}
|
||||
|
||||
size_t DecodePlugin::getSerializationSize() const
|
||||
{
|
||||
return 0;
|
||||
}
|
||||
|
||||
int DecodePlugin::initialize()
|
||||
{
|
||||
return 0;
|
||||
}
|
||||
|
||||
Dims DecodePlugin::getOutputDimensions(int index, const Dims* inputs, int nbInputDims)
|
||||
{
|
||||
//output the result to channel
|
||||
int totalCount = 0;
|
||||
totalCount += decodeplugin::INPUT_H / 8 * decodeplugin::INPUT_W / 8 * 2 * sizeof(decodeplugin::Detection) / sizeof(float);
|
||||
totalCount += decodeplugin::INPUT_H / 16 * decodeplugin::INPUT_W / 16 * 2 * sizeof(decodeplugin::Detection) / sizeof(float);
|
||||
totalCount += decodeplugin::INPUT_H / 32 * decodeplugin::INPUT_W / 32 * 2 * sizeof(decodeplugin::Detection) / sizeof(float);
|
||||
|
||||
return Dims3(totalCount + 1, 1, 1);
|
||||
}
|
||||
|
||||
// Set plugin namespace
|
||||
void DecodePlugin::setPluginNamespace(const char* pluginNamespace)
|
||||
{
|
||||
mPluginNamespace = pluginNamespace;
|
||||
}
|
||||
|
||||
const char* DecodePlugin::getPluginNamespace() const
|
||||
{
|
||||
return mPluginNamespace;
|
||||
}
|
||||
|
||||
// Return the DataType of the plugin output at the requested index
|
||||
DataType DecodePlugin::getOutputDataType(int index, const nvinfer1::DataType* inputTypes, int nbInputs) const
|
||||
{
|
||||
return DataType::kFLOAT;
|
||||
}
|
||||
|
||||
// Return true if output tensor is broadcast across a batch.
|
||||
bool DecodePlugin::isOutputBroadcastAcrossBatch(int outputIndex, const bool* inputIsBroadcasted, int nbInputs) const
|
||||
{
|
||||
return false;
|
||||
}
|
||||
|
||||
// Return true if plugin can use input that is broadcast across batch without replication.
|
||||
bool DecodePlugin::canBroadcastInputAcrossBatch(int inputIndex) const
|
||||
{
|
||||
return false;
|
||||
}
|
||||
|
||||
void DecodePlugin::configurePlugin(const PluginTensorDesc* in, int nbInput, const PluginTensorDesc* out, int nbOutput)
|
||||
{
|
||||
}
|
||||
|
||||
// Attach the plugin object to an execution context and grant the plugin the access to some context resource.
|
||||
void DecodePlugin::attachToContext(cudnnContext* cudnnContext, cublasContext* cublasContext, IGpuAllocator* gpuAllocator)
|
||||
{
|
||||
}
|
||||
|
||||
// Detach the plugin object from its execution context.
|
||||
void DecodePlugin::detachFromContext() {}
|
||||
|
||||
const char* DecodePlugin::getPluginType() const
|
||||
{
|
||||
return "Decode_TRT";
|
||||
}
|
||||
|
||||
const char* DecodePlugin::getPluginVersion() const
|
||||
{
|
||||
return "1";
|
||||
}
|
||||
|
||||
void DecodePlugin::destroy()
|
||||
{
|
||||
delete this;
|
||||
}
|
||||
|
||||
// Clone the plugin
|
||||
IPluginV2IOExt* DecodePlugin::clone() const
|
||||
{
|
||||
DecodePlugin *p = new DecodePlugin();
|
||||
p->setPluginNamespace(mPluginNamespace);
|
||||
return p;
|
||||
}
|
||||
|
||||
__device__ float Logist(float data){ return 1./(1. + expf(-data)); };
|
||||
|
||||
__global__ void CalDetection(const float *input, float *output, int num_elem, int step, int anchor) {
|
||||
|
||||
int idx = threadIdx.x + blockDim.x * blockIdx.x;
|
||||
if (idx >= num_elem) return;
|
||||
|
||||
int h = decodeplugin::INPUT_H / step;
|
||||
int w = decodeplugin::INPUT_W / step;
|
||||
int y = idx / w;
|
||||
int x = idx % w;
|
||||
const float *cls_reg = &input[2 * num_elem];
|
||||
const float *bbox_reg = &input[4 * num_elem];
|
||||
const float *lmk_reg = &input[12 * num_elem];
|
||||
const float *mask_reg = &input[36 * num_elem];
|
||||
|
||||
for (int k = 0; k < 2; ++k) {
|
||||
float conf = cls_reg[idx + k * num_elem];
|
||||
if (conf < 0.5) continue;
|
||||
|
||||
float *res_count = output;
|
||||
int count = (int)atomicAdd(res_count, 1);
|
||||
char* data = (char *)res_count + sizeof(float) + count * sizeof(decodeplugin::Detection);
|
||||
decodeplugin::Detection* det = (decodeplugin::Detection*)(data);
|
||||
|
||||
float prior[4];
|
||||
prior[0] = 7.5 + (float)(x * step);
|
||||
prior[1] = 7.5 + (float)(y * step);
|
||||
prior[2] = anchor * 2 / (k + 1);
|
||||
prior[3] = prior[2];
|
||||
|
||||
//Location
|
||||
det->bbox[0] = prior[0] + bbox_reg[idx + k * num_elem * 4] * prior[2];
|
||||
det->bbox[1] = prior[1] + bbox_reg[idx + k * num_elem * 4 + num_elem] * prior[3];
|
||||
det->bbox[2] = prior[2] * expf(bbox_reg[idx + k * num_elem * 4 + num_elem * 2]);
|
||||
det->bbox[3] = prior[3] * expf(bbox_reg[idx + k * num_elem * 4 + num_elem * 3]);
|
||||
det->bbox[0] -= (det->bbox[2] - 1) / 2;
|
||||
det->bbox[1] -= (det->bbox[3] - 1) / 2;
|
||||
det->bbox[2] += det->bbox[0];
|
||||
det->bbox[3] += det->bbox[1];
|
||||
det->class_confidence = conf;
|
||||
for (int i = 0; i < 10; i += 2) {
|
||||
det->landmark[i] = prior[0] + lmk_reg[idx + k * num_elem * 10 + num_elem * i] * 0.2 * prior[2];
|
||||
det->landmark[i+1] = prior[1] + lmk_reg[idx + k * num_elem * 10 + num_elem * (i + 1)] * 0.2 * prior[3];
|
||||
}
|
||||
det->mask_confidence = mask_reg[idx + k * num_elem];;
|
||||
}
|
||||
}
|
||||
|
||||
void DecodePlugin::forwardGpu(const float *const * inputs, float * output, cudaStream_t stream, int batchSize)
|
||||
{
|
||||
int num_elem = 0;
|
||||
int base_step = 8;
|
||||
int base_anchor = 16;
|
||||
int thread_count;
|
||||
cudaMemset(output, 0, sizeof(float));
|
||||
for (unsigned int i = 0; i < 3; ++i)
|
||||
{
|
||||
num_elem = decodeplugin::INPUT_H / base_step * decodeplugin::INPUT_W / base_step;
|
||||
thread_count = (num_elem < thread_count_) ? num_elem : thread_count_;
|
||||
CalDetection<<< (num_elem + thread_count - 1) / thread_count, thread_count>>>
|
||||
(inputs[i], output, num_elem, base_step, base_anchor);
|
||||
base_step *= 2;
|
||||
base_anchor *= 4;
|
||||
}
|
||||
}
|
||||
|
||||
int DecodePlugin::enqueue(int batchSize, const void*const * inputs, void** outputs, void* workspace, cudaStream_t stream)
|
||||
{
|
||||
//assert(batchSize == 1);
|
||||
//GPU
|
||||
//CUDA_CHECK(cudaStreamSynchronize(stream));
|
||||
forwardGpu((const float *const *)inputs,(float *)outputs[0],stream,batchSize);
|
||||
|
||||
return 0;
|
||||
};
|
||||
|
||||
PluginFieldCollection DecodePluginCreator::mFC{};
|
||||
std::vector<PluginField> DecodePluginCreator::mPluginAttributes;
|
||||
|
||||
DecodePluginCreator::DecodePluginCreator()
|
||||
{
|
||||
mPluginAttributes.clear();
|
||||
|
||||
mFC.nbFields = mPluginAttributes.size();
|
||||
mFC.fields = mPluginAttributes.data();
|
||||
}
|
||||
|
||||
const char* DecodePluginCreator::getPluginName() const
|
||||
{
|
||||
return "Decode_TRT";
|
||||
}
|
||||
|
||||
const char* DecodePluginCreator::getPluginVersion() const
|
||||
{
|
||||
return "1";
|
||||
}
|
||||
|
||||
const PluginFieldCollection* DecodePluginCreator::getFieldNames()
|
||||
{
|
||||
return &mFC;
|
||||
}
|
||||
|
||||
IPluginV2IOExt* DecodePluginCreator::createPlugin(const char* name, const PluginFieldCollection* fc)
|
||||
{
|
||||
DecodePlugin* obj = new DecodePlugin();
|
||||
obj->setPluginNamespace(mNamespace.c_str());
|
||||
return obj;
|
||||
}
|
||||
|
||||
IPluginV2IOExt* DecodePluginCreator::deserializePlugin(const char* name, const void* serialData, size_t serialLength)
|
||||
{
|
||||
// This object will be deleted when the network is destroyed, which will
|
||||
// call PReluPlugin::destroy()
|
||||
DecodePlugin* obj = new DecodePlugin(serialData, serialLength);
|
||||
obj->setPluginNamespace(mNamespace.c_str());
|
||||
return obj;
|
||||
}
|
||||
|
||||
}
|
||||
119
retinafaceAntiCov/decode.h
Normal file
119
retinafaceAntiCov/decode.h
Normal file
@ -0,0 +1,119 @@
|
||||
#ifndef _DECODE_CU_H
|
||||
#define _DECODE_CU_H
|
||||
|
||||
#include <string>
|
||||
#include <vector>
|
||||
#include "NvInfer.h"
|
||||
|
||||
namespace decodeplugin
|
||||
{
|
||||
struct alignas(float) Detection{
|
||||
float bbox[4]; //x1 y1 x2 y2
|
||||
float class_confidence;
|
||||
float landmark[10];
|
||||
float mask_confidence;
|
||||
};
|
||||
static const int INPUT_H = 640;
|
||||
static const int INPUT_W = 640;
|
||||
}
|
||||
|
||||
namespace nvinfer1
|
||||
{
|
||||
class DecodePlugin: public IPluginV2IOExt
|
||||
{
|
||||
public:
|
||||
DecodePlugin();
|
||||
DecodePlugin(const void* data, size_t length);
|
||||
|
||||
~DecodePlugin();
|
||||
|
||||
int getNbOutputs() const override
|
||||
{
|
||||
return 1;
|
||||
}
|
||||
|
||||
Dims getOutputDimensions(int index, const Dims* inputs, int nbInputDims) override;
|
||||
|
||||
int initialize() override;
|
||||
|
||||
virtual void terminate() override {};
|
||||
|
||||
virtual size_t getWorkspaceSize(int maxBatchSize) const override { return 0;}
|
||||
|
||||
virtual int enqueue(int batchSize, const void*const * inputs, void** outputs, void* workspace, cudaStream_t stream) override;
|
||||
|
||||
virtual size_t getSerializationSize() const override;
|
||||
|
||||
virtual void serialize(void* buffer) const override;
|
||||
|
||||
bool supportsFormatCombination(int pos, const PluginTensorDesc* inOut, int nbInputs, int nbOutputs) const override {
|
||||
return inOut[pos].format == TensorFormat::kLINEAR && inOut[pos].type == DataType::kFLOAT;
|
||||
}
|
||||
|
||||
const char* getPluginType() const override;
|
||||
|
||||
const char* getPluginVersion() const override;
|
||||
|
||||
void destroy() override;
|
||||
|
||||
IPluginV2IOExt* clone() const override;
|
||||
|
||||
void setPluginNamespace(const char* pluginNamespace) override;
|
||||
|
||||
const char* getPluginNamespace() const override;
|
||||
|
||||
DataType getOutputDataType(int index, const nvinfer1::DataType* inputTypes, int nbInputs) const override;
|
||||
|
||||
bool isOutputBroadcastAcrossBatch(int outputIndex, const bool* inputIsBroadcasted, int nbInputs) const override;
|
||||
|
||||
bool canBroadcastInputAcrossBatch(int inputIndex) const override;
|
||||
|
||||
void attachToContext(
|
||||
cudnnContext* cudnnContext, cublasContext* cublasContext, IGpuAllocator* gpuAllocator) override;
|
||||
|
||||
void configurePlugin(const PluginTensorDesc* in, int nbInput, const PluginTensorDesc* out, int nbOutput) override;
|
||||
|
||||
void detachFromContext() override;
|
||||
|
||||
int input_size_;
|
||||
private:
|
||||
void forwardGpu(const float *const * inputs, float* output, cudaStream_t stream, int batchSize = 1);
|
||||
int thread_count_ = 256;
|
||||
const char* mPluginNamespace;
|
||||
};
|
||||
|
||||
class DecodePluginCreator : public IPluginCreator
|
||||
{
|
||||
public:
|
||||
DecodePluginCreator();
|
||||
|
||||
~DecodePluginCreator() override = default;
|
||||
|
||||
const char* getPluginName() const override;
|
||||
|
||||
const char* getPluginVersion() const override;
|
||||
|
||||
const PluginFieldCollection* getFieldNames() override;
|
||||
|
||||
IPluginV2IOExt* createPlugin(const char* name, const PluginFieldCollection* fc) override;
|
||||
|
||||
IPluginV2IOExt* deserializePlugin(const char* name, const void* serialData, size_t serialLength) override;
|
||||
|
||||
void setPluginNamespace(const char* libNamespace) override
|
||||
{
|
||||
mNamespace = libNamespace;
|
||||
}
|
||||
|
||||
const char* getPluginNamespace() const override
|
||||
{
|
||||
return mNamespace.c_str();
|
||||
}
|
||||
|
||||
private:
|
||||
std::string mNamespace;
|
||||
static PluginFieldCollection mFC;
|
||||
static std::vector<PluginField> mPluginAttributes;
|
||||
};
|
||||
};
|
||||
|
||||
#endif
|
||||
503
retinafaceAntiCov/logging.h
Normal file
503
retinafaceAntiCov/logging.h
Normal file
@ -0,0 +1,503 @@
|
||||
/*
|
||||
* Copyright (c) 2019, NVIDIA CORPORATION. All rights reserved.
|
||||
*
|
||||
* Licensed under the Apache License, Version 2.0 (the "License");
|
||||
* you may not use this file except in compliance with the License.
|
||||
* You may obtain a copy of the License at
|
||||
*
|
||||
* http://www.apache.org/licenses/LICENSE-2.0
|
||||
*
|
||||
* Unless required by applicable law or agreed to in writing, software
|
||||
* distributed under the License is distributed on an "AS IS" BASIS,
|
||||
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
||||
* See the License for the specific language governing permissions and
|
||||
* limitations under the License.
|
||||
*/
|
||||
|
||||
#ifndef TENSORRT_LOGGING_H
|
||||
#define TENSORRT_LOGGING_H
|
||||
|
||||
#include "NvInferRuntimeCommon.h"
|
||||
#include <cassert>
|
||||
#include <ctime>
|
||||
#include <iomanip>
|
||||
#include <iostream>
|
||||
#include <ostream>
|
||||
#include <sstream>
|
||||
#include <string>
|
||||
|
||||
using Severity = nvinfer1::ILogger::Severity;
|
||||
|
||||
class LogStreamConsumerBuffer : public std::stringbuf
|
||||
{
|
||||
public:
|
||||
LogStreamConsumerBuffer(std::ostream& stream, const std::string& prefix, bool shouldLog)
|
||||
: mOutput(stream)
|
||||
, mPrefix(prefix)
|
||||
, mShouldLog(shouldLog)
|
||||
{
|
||||
}
|
||||
|
||||
LogStreamConsumerBuffer(LogStreamConsumerBuffer&& other)
|
||||
: mOutput(other.mOutput)
|
||||
{
|
||||
}
|
||||
|
||||
~LogStreamConsumerBuffer()
|
||||
{
|
||||
// std::streambuf::pbase() gives a pointer to the beginning of the buffered part of the output sequence
|
||||
// std::streambuf::pptr() gives a pointer to the current position of the output sequence
|
||||
// if the pointer to the beginning is not equal to the pointer to the current position,
|
||||
// call putOutput() to log the output to the stream
|
||||
if (pbase() != pptr())
|
||||
{
|
||||
putOutput();
|
||||
}
|
||||
}
|
||||
|
||||
// synchronizes the stream buffer and returns 0 on success
|
||||
// synchronizing the stream buffer consists of inserting the buffer contents into the stream,
|
||||
// resetting the buffer and flushing the stream
|
||||
virtual int sync()
|
||||
{
|
||||
putOutput();
|
||||
return 0;
|
||||
}
|
||||
|
||||
void putOutput()
|
||||
{
|
||||
if (mShouldLog)
|
||||
{
|
||||
// prepend timestamp
|
||||
std::time_t timestamp = std::time(nullptr);
|
||||
tm* tm_local = std::localtime(×tamp);
|
||||
std::cout << "[";
|
||||
std::cout << std::setw(2) << std::setfill('0') << 1 + tm_local->tm_mon << "/";
|
||||
std::cout << std::setw(2) << std::setfill('0') << tm_local->tm_mday << "/";
|
||||
std::cout << std::setw(4) << std::setfill('0') << 1900 + tm_local->tm_year << "-";
|
||||
std::cout << std::setw(2) << std::setfill('0') << tm_local->tm_hour << ":";
|
||||
std::cout << std::setw(2) << std::setfill('0') << tm_local->tm_min << ":";
|
||||
std::cout << std::setw(2) << std::setfill('0') << tm_local->tm_sec << "] ";
|
||||
// std::stringbuf::str() gets the string contents of the buffer
|
||||
// insert the buffer contents pre-appended by the appropriate prefix into the stream
|
||||
mOutput << mPrefix << str();
|
||||
// set the buffer to empty
|
||||
str("");
|
||||
// flush the stream
|
||||
mOutput.flush();
|
||||
}
|
||||
}
|
||||
|
||||
void setShouldLog(bool shouldLog)
|
||||
{
|
||||
mShouldLog = shouldLog;
|
||||
}
|
||||
|
||||
private:
|
||||
std::ostream& mOutput;
|
||||
std::string mPrefix;
|
||||
bool mShouldLog;
|
||||
};
|
||||
|
||||
//!
|
||||
//! \class LogStreamConsumerBase
|
||||
//! \brief Convenience object used to initialize LogStreamConsumerBuffer before std::ostream in LogStreamConsumer
|
||||
//!
|
||||
class LogStreamConsumerBase
|
||||
{
|
||||
public:
|
||||
LogStreamConsumerBase(std::ostream& stream, const std::string& prefix, bool shouldLog)
|
||||
: mBuffer(stream, prefix, shouldLog)
|
||||
{
|
||||
}
|
||||
|
||||
protected:
|
||||
LogStreamConsumerBuffer mBuffer;
|
||||
};
|
||||
|
||||
//!
|
||||
//! \class LogStreamConsumer
|
||||
//! \brief Convenience object used to facilitate use of C++ stream syntax when logging messages.
|
||||
//! Order of base classes is LogStreamConsumerBase and then std::ostream.
|
||||
//! This is because the LogStreamConsumerBase class is used to initialize the LogStreamConsumerBuffer member field
|
||||
//! in LogStreamConsumer and then the address of the buffer is passed to std::ostream.
|
||||
//! This is necessary to prevent the address of an uninitialized buffer from being passed to std::ostream.
|
||||
//! Please do not change the order of the parent classes.
|
||||
//!
|
||||
class LogStreamConsumer : protected LogStreamConsumerBase, public std::ostream
|
||||
{
|
||||
public:
|
||||
//! \brief Creates a LogStreamConsumer which logs messages with level severity.
|
||||
//! Reportable severity determines if the messages are severe enough to be logged.
|
||||
LogStreamConsumer(Severity reportableSeverity, Severity severity)
|
||||
: LogStreamConsumerBase(severityOstream(severity), severityPrefix(severity), severity <= reportableSeverity)
|
||||
, std::ostream(&mBuffer) // links the stream buffer with the stream
|
||||
, mShouldLog(severity <= reportableSeverity)
|
||||
, mSeverity(severity)
|
||||
{
|
||||
}
|
||||
|
||||
LogStreamConsumer(LogStreamConsumer&& other)
|
||||
: LogStreamConsumerBase(severityOstream(other.mSeverity), severityPrefix(other.mSeverity), other.mShouldLog)
|
||||
, std::ostream(&mBuffer) // links the stream buffer with the stream
|
||||
, mShouldLog(other.mShouldLog)
|
||||
, mSeverity(other.mSeverity)
|
||||
{
|
||||
}
|
||||
|
||||
void setReportableSeverity(Severity reportableSeverity)
|
||||
{
|
||||
mShouldLog = mSeverity <= reportableSeverity;
|
||||
mBuffer.setShouldLog(mShouldLog);
|
||||
}
|
||||
|
||||
private:
|
||||
static std::ostream& severityOstream(Severity severity)
|
||||
{
|
||||
return severity >= Severity::kINFO ? std::cout : std::cerr;
|
||||
}
|
||||
|
||||
static std::string severityPrefix(Severity severity)
|
||||
{
|
||||
switch (severity)
|
||||
{
|
||||
case Severity::kINTERNAL_ERROR: return "[F] ";
|
||||
case Severity::kERROR: return "[E] ";
|
||||
case Severity::kWARNING: return "[W] ";
|
||||
case Severity::kINFO: return "[I] ";
|
||||
case Severity::kVERBOSE: return "[V] ";
|
||||
default: assert(0); return "";
|
||||
}
|
||||
}
|
||||
|
||||
bool mShouldLog;
|
||||
Severity mSeverity;
|
||||
};
|
||||
|
||||
//! \class Logger
|
||||
//!
|
||||
//! \brief Class which manages logging of TensorRT tools and samples
|
||||
//!
|
||||
//! \details This class provides a common interface for TensorRT tools and samples to log information to the console,
|
||||
//! and supports logging two types of messages:
|
||||
//!
|
||||
//! - Debugging messages with an associated severity (info, warning, error, or internal error/fatal)
|
||||
//! - Test pass/fail messages
|
||||
//!
|
||||
//! The advantage of having all samples use this class for logging as opposed to emitting directly to stdout/stderr is
|
||||
//! that the logic for controlling the verbosity and formatting of sample output is centralized in one location.
|
||||
//!
|
||||
//! In the future, this class could be extended to support dumping test results to a file in some standard format
|
||||
//! (for example, JUnit XML), and providing additional metadata (e.g. timing the duration of a test run).
|
||||
//!
|
||||
//! TODO: For backwards compatibility with existing samples, this class inherits directly from the nvinfer1::ILogger
|
||||
//! interface, which is problematic since there isn't a clean separation between messages coming from the TensorRT
|
||||
//! library and messages coming from the sample.
|
||||
//!
|
||||
//! In the future (once all samples are updated to use Logger::getTRTLogger() to access the ILogger) we can refactor the
|
||||
//! class to eliminate the inheritance and instead make the nvinfer1::ILogger implementation a member of the Logger
|
||||
//! object.
|
||||
|
||||
class Logger : public nvinfer1::ILogger
|
||||
{
|
||||
public:
|
||||
Logger(Severity severity = Severity::kWARNING)
|
||||
: mReportableSeverity(severity)
|
||||
{
|
||||
}
|
||||
|
||||
//!
|
||||
//! \enum TestResult
|
||||
//! \brief Represents the state of a given test
|
||||
//!
|
||||
enum class TestResult
|
||||
{
|
||||
kRUNNING, //!< The test is running
|
||||
kPASSED, //!< The test passed
|
||||
kFAILED, //!< The test failed
|
||||
kWAIVED //!< The test was waived
|
||||
};
|
||||
|
||||
//!
|
||||
//! \brief Forward-compatible method for retrieving the nvinfer::ILogger associated with this Logger
|
||||
//! \return The nvinfer1::ILogger associated with this Logger
|
||||
//!
|
||||
//! TODO Once all samples are updated to use this method to register the logger with TensorRT,
|
||||
//! we can eliminate the inheritance of Logger from ILogger
|
||||
//!
|
||||
nvinfer1::ILogger& getTRTLogger()
|
||||
{
|
||||
return *this;
|
||||
}
|
||||
|
||||
//!
|
||||
//! \brief Implementation of the nvinfer1::ILogger::log() virtual method
|
||||
//!
|
||||
//! Note samples should not be calling this function directly; it will eventually go away once we eliminate the
|
||||
//! inheritance from nvinfer1::ILogger
|
||||
//!
|
||||
void log(Severity severity, const char* msg) override
|
||||
{
|
||||
LogStreamConsumer(mReportableSeverity, severity) << "[TRT] " << std::string(msg) << std::endl;
|
||||
}
|
||||
|
||||
//!
|
||||
//! \brief Method for controlling the verbosity of logging output
|
||||
//!
|
||||
//! \param severity The logger will only emit messages that have severity of this level or higher.
|
||||
//!
|
||||
void setReportableSeverity(Severity severity)
|
||||
{
|
||||
mReportableSeverity = severity;
|
||||
}
|
||||
|
||||
//!
|
||||
//! \brief Opaque handle that holds logging information for a particular test
|
||||
//!
|
||||
//! This object is an opaque handle to information used by the Logger to print test results.
|
||||
//! The sample must call Logger::defineTest() in order to obtain a TestAtom that can be used
|
||||
//! with Logger::reportTest{Start,End}().
|
||||
//!
|
||||
class TestAtom
|
||||
{
|
||||
public:
|
||||
TestAtom(TestAtom&&) = default;
|
||||
|
||||
private:
|
||||
friend class Logger;
|
||||
|
||||
TestAtom(bool started, const std::string& name, const std::string& cmdline)
|
||||
: mStarted(started)
|
||||
, mName(name)
|
||||
, mCmdline(cmdline)
|
||||
{
|
||||
}
|
||||
|
||||
bool mStarted;
|
||||
std::string mName;
|
||||
std::string mCmdline;
|
||||
};
|
||||
|
||||
//!
|
||||
//! \brief Define a test for logging
|
||||
//!
|
||||
//! \param[in] name The name of the test. This should be a string starting with
|
||||
//! "TensorRT" and containing dot-separated strings containing
|
||||
//! the characters [A-Za-z0-9_].
|
||||
//! For example, "TensorRT.sample_googlenet"
|
||||
//! \param[in] cmdline The command line used to reproduce the test
|
||||
//
|
||||
//! \return a TestAtom that can be used in Logger::reportTest{Start,End}().
|
||||
//!
|
||||
static TestAtom defineTest(const std::string& name, const std::string& cmdline)
|
||||
{
|
||||
return TestAtom(false, name, cmdline);
|
||||
}
|
||||
|
||||
//!
|
||||
//! \brief A convenience overloaded version of defineTest() that accepts an array of command-line arguments
|
||||
//! as input
|
||||
//!
|
||||
//! \param[in] name The name of the test
|
||||
//! \param[in] argc The number of command-line arguments
|
||||
//! \param[in] argv The array of command-line arguments (given as C strings)
|
||||
//!
|
||||
//! \return a TestAtom that can be used in Logger::reportTest{Start,End}().
|
||||
static TestAtom defineTest(const std::string& name, int argc, char const* const* argv)
|
||||
{
|
||||
auto cmdline = genCmdlineString(argc, argv);
|
||||
return defineTest(name, cmdline);
|
||||
}
|
||||
|
||||
//!
|
||||
//! \brief Report that a test has started.
|
||||
//!
|
||||
//! \pre reportTestStart() has not been called yet for the given testAtom
|
||||
//!
|
||||
//! \param[in] testAtom The handle to the test that has started
|
||||
//!
|
||||
static void reportTestStart(TestAtom& testAtom)
|
||||
{
|
||||
reportTestResult(testAtom, TestResult::kRUNNING);
|
||||
assert(!testAtom.mStarted);
|
||||
testAtom.mStarted = true;
|
||||
}
|
||||
|
||||
//!
|
||||
//! \brief Report that a test has ended.
|
||||
//!
|
||||
//! \pre reportTestStart() has been called for the given testAtom
|
||||
//!
|
||||
//! \param[in] testAtom The handle to the test that has ended
|
||||
//! \param[in] result The result of the test. Should be one of TestResult::kPASSED,
|
||||
//! TestResult::kFAILED, TestResult::kWAIVED
|
||||
//!
|
||||
static void reportTestEnd(const TestAtom& testAtom, TestResult result)
|
||||
{
|
||||
assert(result != TestResult::kRUNNING);
|
||||
assert(testAtom.mStarted);
|
||||
reportTestResult(testAtom, result);
|
||||
}
|
||||
|
||||
static int reportPass(const TestAtom& testAtom)
|
||||
{
|
||||
reportTestEnd(testAtom, TestResult::kPASSED);
|
||||
return EXIT_SUCCESS;
|
||||
}
|
||||
|
||||
static int reportFail(const TestAtom& testAtom)
|
||||
{
|
||||
reportTestEnd(testAtom, TestResult::kFAILED);
|
||||
return EXIT_FAILURE;
|
||||
}
|
||||
|
||||
static int reportWaive(const TestAtom& testAtom)
|
||||
{
|
||||
reportTestEnd(testAtom, TestResult::kWAIVED);
|
||||
return EXIT_SUCCESS;
|
||||
}
|
||||
|
||||
static int reportTest(const TestAtom& testAtom, bool pass)
|
||||
{
|
||||
return pass ? reportPass(testAtom) : reportFail(testAtom);
|
||||
}
|
||||
|
||||
Severity getReportableSeverity() const
|
||||
{
|
||||
return mReportableSeverity;
|
||||
}
|
||||
|
||||
private:
|
||||
//!
|
||||
//! \brief returns an appropriate string for prefixing a log message with the given severity
|
||||
//!
|
||||
static const char* severityPrefix(Severity severity)
|
||||
{
|
||||
switch (severity)
|
||||
{
|
||||
case Severity::kINTERNAL_ERROR: return "[F] ";
|
||||
case Severity::kERROR: return "[E] ";
|
||||
case Severity::kWARNING: return "[W] ";
|
||||
case Severity::kINFO: return "[I] ";
|
||||
case Severity::kVERBOSE: return "[V] ";
|
||||
default: assert(0); return "";
|
||||
}
|
||||
}
|
||||
|
||||
//!
|
||||
//! \brief returns an appropriate string for prefixing a test result message with the given result
|
||||
//!
|
||||
static const char* testResultString(TestResult result)
|
||||
{
|
||||
switch (result)
|
||||
{
|
||||
case TestResult::kRUNNING: return "RUNNING";
|
||||
case TestResult::kPASSED: return "PASSED";
|
||||
case TestResult::kFAILED: return "FAILED";
|
||||
case TestResult::kWAIVED: return "WAIVED";
|
||||
default: assert(0); return "";
|
||||
}
|
||||
}
|
||||
|
||||
//!
|
||||
//! \brief returns an appropriate output stream (cout or cerr) to use with the given severity
|
||||
//!
|
||||
static std::ostream& severityOstream(Severity severity)
|
||||
{
|
||||
return severity >= Severity::kINFO ? std::cout : std::cerr;
|
||||
}
|
||||
|
||||
//!
|
||||
//! \brief method that implements logging test results
|
||||
//!
|
||||
static void reportTestResult(const TestAtom& testAtom, TestResult result)
|
||||
{
|
||||
severityOstream(Severity::kINFO) << "&&&& " << testResultString(result) << " " << testAtom.mName << " # "
|
||||
<< testAtom.mCmdline << std::endl;
|
||||
}
|
||||
|
||||
//!
|
||||
//! \brief generate a command line string from the given (argc, argv) values
|
||||
//!
|
||||
static std::string genCmdlineString(int argc, char const* const* argv)
|
||||
{
|
||||
std::stringstream ss;
|
||||
for (int i = 0; i < argc; i++)
|
||||
{
|
||||
if (i > 0)
|
||||
ss << " ";
|
||||
ss << argv[i];
|
||||
}
|
||||
return ss.str();
|
||||
}
|
||||
|
||||
Severity mReportableSeverity;
|
||||
};
|
||||
|
||||
namespace
|
||||
{
|
||||
|
||||
//!
|
||||
//! \brief produces a LogStreamConsumer object that can be used to log messages of severity kVERBOSE
|
||||
//!
|
||||
//! Example usage:
|
||||
//!
|
||||
//! LOG_VERBOSE(logger) << "hello world" << std::endl;
|
||||
//!
|
||||
inline LogStreamConsumer LOG_VERBOSE(const Logger& logger)
|
||||
{
|
||||
return LogStreamConsumer(logger.getReportableSeverity(), Severity::kVERBOSE);
|
||||
}
|
||||
|
||||
//!
|
||||
//! \brief produces a LogStreamConsumer object that can be used to log messages of severity kINFO
|
||||
//!
|
||||
//! Example usage:
|
||||
//!
|
||||
//! LOG_INFO(logger) << "hello world" << std::endl;
|
||||
//!
|
||||
inline LogStreamConsumer LOG_INFO(const Logger& logger)
|
||||
{
|
||||
return LogStreamConsumer(logger.getReportableSeverity(), Severity::kINFO);
|
||||
}
|
||||
|
||||
//!
|
||||
//! \brief produces a LogStreamConsumer object that can be used to log messages of severity kWARNING
|
||||
//!
|
||||
//! Example usage:
|
||||
//!
|
||||
//! LOG_WARN(logger) << "hello world" << std::endl;
|
||||
//!
|
||||
inline LogStreamConsumer LOG_WARN(const Logger& logger)
|
||||
{
|
||||
return LogStreamConsumer(logger.getReportableSeverity(), Severity::kWARNING);
|
||||
}
|
||||
|
||||
//!
|
||||
//! \brief produces a LogStreamConsumer object that can be used to log messages of severity kERROR
|
||||
//!
|
||||
//! Example usage:
|
||||
//!
|
||||
//! LOG_ERROR(logger) << "hello world" << std::endl;
|
||||
//!
|
||||
inline LogStreamConsumer LOG_ERROR(const Logger& logger)
|
||||
{
|
||||
return LogStreamConsumer(logger.getReportableSeverity(), Severity::kERROR);
|
||||
}
|
||||
|
||||
//!
|
||||
//! \brief produces a LogStreamConsumer object that can be used to log messages of severity kINTERNAL_ERROR
|
||||
// ("fatal" severity)
|
||||
//!
|
||||
//! Example usage:
|
||||
//!
|
||||
//! LOG_FATAL(logger) << "hello world" << std::endl;
|
||||
//!
|
||||
inline LogStreamConsumer LOG_FATAL(const Logger& logger)
|
||||
{
|
||||
return LogStreamConsumer(logger.getReportableSeverity(), Severity::kINTERNAL_ERROR);
|
||||
}
|
||||
|
||||
} // anonymous namespace
|
||||
|
||||
#endif // TENSORRT_LOGGING_H
|
||||
570
retinafaceAntiCov/retinafaceAntiCov.cpp
Normal file
570
retinafaceAntiCov/retinafaceAntiCov.cpp
Normal file
@ -0,0 +1,570 @@
|
||||
#include <fstream>
|
||||
#include <iostream>
|
||||
#include <map>
|
||||
#include <sstream>
|
||||
#include <vector>
|
||||
#include <chrono>
|
||||
#include <opencv2/opencv.hpp>
|
||||
#include <dirent.h>
|
||||
#include "NvInfer.h"
|
||||
#include "cuda_runtime_api.h"
|
||||
#include "logging.h"
|
||||
#include "decode.h"
|
||||
|
||||
#define CHECK(status) \
|
||||
do\
|
||||
{\
|
||||
auto ret = (status);\
|
||||
if (ret != 0)\
|
||||
{\
|
||||
std::cerr << "Cuda failure: " << ret << std::endl;\
|
||||
abort();\
|
||||
}\
|
||||
} while (0)
|
||||
|
||||
//#define USE_FP16 // comment out this if want to use FP32
|
||||
#define DEVICE 0 // GPU id
|
||||
#define BATCH_SIZE 1 // currently, only support BATCH=1
|
||||
|
||||
using namespace nvinfer1;
|
||||
|
||||
// stuff we know about the network and the input/output blobs
|
||||
static const int INPUT_H = 640;
|
||||
static const int INPUT_W = 640;
|
||||
static const int DETECTION_SIZE = sizeof(decodeplugin::Detection) / sizeof(float);
|
||||
static const int OUTPUT_SIZE = (INPUT_H / 8 * INPUT_W / 8 + INPUT_H / 16 * INPUT_W / 16 + INPUT_H / 32 * INPUT_W / 32) * 2 * DETECTION_SIZE + 1;
|
||||
const char* INPUT_BLOB_NAME = "data";
|
||||
const char* OUTPUT_BLOB_NAME = "prob";
|
||||
static Logger gLogger;
|
||||
REGISTER_TENSORRT_PLUGIN(DecodePluginCreator);
|
||||
|
||||
cv::Mat preprocess_img(cv::Mat& img) {
|
||||
int w, h, x, y;
|
||||
float r_w = INPUT_W / (img.cols*1.0);
|
||||
float r_h = INPUT_H / (img.rows*1.0);
|
||||
if (r_h > r_w) {
|
||||
w = INPUT_W;
|
||||
h = r_w * img.rows;
|
||||
x = 0;
|
||||
y = (INPUT_H - h) / 2;
|
||||
} else {
|
||||
w = r_h* img.cols;
|
||||
h = INPUT_H;
|
||||
x = (INPUT_W - w) / 2;
|
||||
y = 0;
|
||||
}
|
||||
cv::Mat re(h, w, CV_8UC3);
|
||||
cv::resize(img, re, re.size(), 0, 0, cv::INTER_CUBIC);
|
||||
cv::Mat out(INPUT_H, INPUT_W, CV_8UC3, cv::Scalar(128, 128, 128));
|
||||
re.copyTo(out(cv::Rect(x, y, re.cols, re.rows)));
|
||||
return out;
|
||||
}
|
||||
|
||||
cv::Rect get_rect_adapt_landmark(cv::Mat& img, float bbox[4], float lmk[10]) {
|
||||
int l, r, t, b;
|
||||
float r_w = INPUT_W / (img.cols * 1.0);
|
||||
float r_h = INPUT_H / (img.rows * 1.0);
|
||||
if (r_h > r_w) {
|
||||
l = bbox[0] / r_w;
|
||||
r = bbox[2] / r_w;
|
||||
t = (bbox[1] - (INPUT_H - r_w * img.rows) / 2) / r_w;
|
||||
b = (bbox[3] - (INPUT_H - r_w * img.rows) / 2) / r_w;
|
||||
for (int i = 0; i < 10; i += 2) {
|
||||
lmk[i] /= r_w;
|
||||
lmk[i + 1] = (lmk[i + 1] - (INPUT_H - r_w * img.rows) / 2) / r_w;
|
||||
}
|
||||
} else {
|
||||
l = (bbox[0] - (INPUT_W - r_h * img.cols) / 2) / r_h;
|
||||
r = (bbox[2] - (INPUT_W - r_h * img.cols) / 2) / r_h;
|
||||
t = bbox[1] / r_h;
|
||||
b = bbox[3] / r_h;
|
||||
for (int i = 0; i < 10; i += 2) {
|
||||
lmk[i] = (lmk[i] - (INPUT_W - r_h * img.cols) / 2) / r_h;
|
||||
lmk[i + 1] /= r_h;
|
||||
}
|
||||
}
|
||||
return cv::Rect(l, t, r-l, b-t);
|
||||
}
|
||||
|
||||
float iou(float lbox[4], float rbox[4]) {
|
||||
float interBox[] = {
|
||||
std::max(lbox[0], rbox[0]), //left
|
||||
std::min(lbox[2], rbox[2]), //right
|
||||
std::max(lbox[1], rbox[1]), //top
|
||||
std::min(lbox[3], rbox[3]), //bottom
|
||||
};
|
||||
|
||||
if(interBox[2] > interBox[3] || interBox[0] > interBox[1])
|
||||
return 0.0f;
|
||||
|
||||
float interBoxS = (interBox[1] - interBox[0]) * (interBox[3] - interBox[2]);
|
||||
return interBoxS / ((lbox[2] - lbox[0]) * (lbox[3] - lbox[1]) + (rbox[2] - rbox[0]) * (rbox[3] - rbox[1]) -interBoxS + 0.000001f);
|
||||
}
|
||||
|
||||
bool cmp(decodeplugin::Detection& a, decodeplugin::Detection& b) {
|
||||
return a.class_confidence > b.class_confidence;
|
||||
}
|
||||
|
||||
void nms(std::vector<decodeplugin::Detection>& res, float *output, float nms_thresh = 0.4) {
|
||||
std::vector<decodeplugin::Detection> dets;
|
||||
for (int i = 0; i < output[0]; i++) {
|
||||
if (output[DETECTION_SIZE * i + 1 + 4] <= 0.1) continue;
|
||||
decodeplugin::Detection det;
|
||||
memcpy(&det, &output[DETECTION_SIZE * i + 1], sizeof(decodeplugin::Detection));
|
||||
dets.push_back(det);
|
||||
}
|
||||
std::sort(dets.begin(), dets.end(), cmp);
|
||||
if (dets.size() > 5000) dets.erase(dets.begin() + 5000, dets.end());
|
||||
for (size_t m = 0; m < dets.size(); ++m) {
|
||||
auto& item = dets[m];
|
||||
res.push_back(item);
|
||||
//std::cout << item.class_confidence << " bbox " << item.bbox[0] << ", " << item.bbox[1] << ", " << item.bbox[2] << ", " << item.bbox[3] << std::endl;
|
||||
for (size_t n = m + 1; n < dets.size(); ++n) {
|
||||
if (iou(item.bbox, dets[n].bbox) > nms_thresh) {
|
||||
dets.erase(dets.begin()+n);
|
||||
--n;
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
// TensorRT weight files have a simple space delimited format:
|
||||
// [type] [size] <data x size in hex>
|
||||
std::map<std::string, Weights> loadWeights(const std::string file) {
|
||||
std::cout << "Loading weights: " << file << std::endl;
|
||||
std::map<std::string, Weights> weightMap;
|
||||
|
||||
// Open weights file
|
||||
std::ifstream input(file);
|
||||
assert(input.is_open() && "Unable to load weight file.");
|
||||
|
||||
// Read number of weight blobs
|
||||
int32_t count;
|
||||
input >> count;
|
||||
assert(count > 0 && "Invalid weight map file.");
|
||||
|
||||
while (count--)
|
||||
{
|
||||
Weights wt{DataType::kFLOAT, nullptr, 0};
|
||||
uint32_t size;
|
||||
|
||||
// Read name and type of blob
|
||||
std::string name;
|
||||
input >> name >> std::dec >> size;
|
||||
wt.type = DataType::kFLOAT;
|
||||
|
||||
// Load blob
|
||||
uint32_t* val = reinterpret_cast<uint32_t*>(malloc(sizeof(val) * size));
|
||||
for (uint32_t x = 0, y = size; x < y; ++x)
|
||||
{
|
||||
input >> std::hex >> val[x];
|
||||
}
|
||||
wt.values = val;
|
||||
|
||||
wt.count = size;
|
||||
weightMap[name] = wt;
|
||||
}
|
||||
|
||||
return weightMap;
|
||||
}
|
||||
|
||||
IScaleLayer* addBatchNorm2d(INetworkDefinition *network, std::map<std::string, Weights>& weightMap, ITensor& input, std::string lname, float eps) {
|
||||
float *gamma = (float*)weightMap[lname + "_gamma"].values;
|
||||
float *beta = (float*)weightMap[lname + "_beta"].values;
|
||||
float *mean = (float*)weightMap[lname + "_moving_mean"].values;
|
||||
float *var = (float*)weightMap[lname + "_moving_var"].values;
|
||||
int len = weightMap[lname + "_moving_var"].count;
|
||||
|
||||
float *scval = reinterpret_cast<float*>(malloc(sizeof(float) * len));
|
||||
for (int i = 0; i < len; i++) {
|
||||
scval[i] = gamma[i] / sqrt(var[i] + eps);
|
||||
}
|
||||
Weights scale{DataType::kFLOAT, scval, len};
|
||||
|
||||
float *shval = reinterpret_cast<float*>(malloc(sizeof(float) * len));
|
||||
for (int i = 0; i < len; i++) {
|
||||
shval[i] = beta[i] - mean[i] * gamma[i] / sqrt(var[i] + eps);
|
||||
}
|
||||
Weights shift{DataType::kFLOAT, shval, len};
|
||||
|
||||
float *pval = reinterpret_cast<float*>(malloc(sizeof(float) * len));
|
||||
for (int i = 0; i < len; i++) {
|
||||
pval[i] = 1.0;
|
||||
}
|
||||
Weights power{DataType::kFLOAT, pval, len};
|
||||
|
||||
weightMap[lname + ".scale"] = scale;
|
||||
weightMap[lname + ".shift"] = shift;
|
||||
weightMap[lname + ".power"] = power;
|
||||
IScaleLayer* scale_1 = network->addScale(input, ScaleMode::kCHANNEL, shift, scale, power);
|
||||
assert(scale_1);
|
||||
return scale_1;
|
||||
}
|
||||
|
||||
ILayer* convBnRelu(INetworkDefinition *network, std::map<std::string, Weights>& weightMap, ITensor& input, int num_filters, int k, int s, int p, int g, std::string lname) {
|
||||
Weights emptywts{DataType::kFLOAT, nullptr, 0};
|
||||
IConvolutionLayer* conv = network->addConvolutionNd(input, num_filters, DimsHW{k, k}, weightMap[lname + "_conv2d_weight"], emptywts);
|
||||
assert(conv);
|
||||
conv->setStrideNd(DimsHW{s, s});
|
||||
conv->setPaddingNd(DimsHW{p, p});
|
||||
conv->setNbGroups(g);
|
||||
auto bn = addBatchNorm2d(network, weightMap, *conv->getOutput(0), lname + "_batchnorm", 1e-3);
|
||||
IActivationLayer* relu = network->addActivation(*bn->getOutput(0), ActivationType::kRELU);
|
||||
assert(relu);
|
||||
return relu;
|
||||
}
|
||||
|
||||
ILayer* convBiasBnRelu(INetworkDefinition *network, std::map<std::string, Weights>& weightMap, ITensor& input, int num_filters, int k, int s, int p, std::string lname) {
|
||||
IConvolutionLayer* conv = network->addConvolutionNd(input, num_filters, DimsHW{k, k}, weightMap[lname + "_weight"], weightMap[lname + "_bias"]);
|
||||
assert(conv);
|
||||
conv->setStrideNd(DimsHW{s, s});
|
||||
conv->setPaddingNd(DimsHW{p, p});
|
||||
auto bn = addBatchNorm2d(network, weightMap, *conv->getOutput(0), lname + "_bn", 2e-5);
|
||||
IActivationLayer* relu = network->addActivation(*bn->getOutput(0), ActivationType::kRELU);
|
||||
assert(relu);
|
||||
return relu;
|
||||
}
|
||||
|
||||
ILayer* head(INetworkDefinition *network, std::map<std::string, Weights>& weightMap, ITensor& input, std::string lname) {
|
||||
auto conv1 = network->addConvolutionNd(input, 32, DimsHW{3, 3}, weightMap[lname + "_conv1_weight"], weightMap[lname + "_conv1_bias"]);
|
||||
assert(conv1);
|
||||
conv1->setPaddingNd(DimsHW{1, 1});
|
||||
auto conv1bn = addBatchNorm2d(network, weightMap, *conv1->getOutput(0), lname + "_conv1_bn", 2e-5);
|
||||
|
||||
auto ctxconv1 = convBiasBnRelu(network, weightMap, input, 16, 3, 1, 1, lname + "_context_conv1");
|
||||
|
||||
auto ctxconv2 = network->addConvolutionNd(*ctxconv1->getOutput(0), 16, DimsHW{3, 3}, weightMap[lname + "_context_conv2_weight"], weightMap[lname + "_context_conv2_bias"]);
|
||||
assert(ctxconv2);
|
||||
ctxconv2->setPaddingNd(DimsHW{1, 1});
|
||||
auto ctxconv2bn = addBatchNorm2d(network, weightMap, *ctxconv2->getOutput(0), lname + "_context_conv2_bn", 2e-5);
|
||||
|
||||
auto ctxconv3_1 = convBiasBnRelu(network, weightMap, *ctxconv1->getOutput(0), 16, 3, 1, 1, lname + "_context_conv3_1");
|
||||
auto ctxconv3_2 = network->addConvolutionNd(*ctxconv3_1->getOutput(0), 16, DimsHW{3, 3}, weightMap[lname + "_context_conv3_2_weight"], weightMap[lname + "_context_conv3_2_bias"]);
|
||||
assert(ctxconv3_2);
|
||||
ctxconv3_2->setPaddingNd(DimsHW{1, 1});
|
||||
auto ctxconv3_2bn = addBatchNorm2d(network, weightMap, *ctxconv3_2->getOutput(0), lname + "_context_conv3_2_bn", 2e-5);
|
||||
|
||||
ITensor* inputTensors[] = {conv1bn->getOutput(0), ctxconv2bn->getOutput(0), ctxconv3_2bn->getOutput(0)};
|
||||
auto cat = network->addConcatenation(inputTensors, 3);
|
||||
assert(cat);
|
||||
|
||||
IActivationLayer* relu = network->addActivation(*cat->getOutput(0), ActivationType::kRELU);
|
||||
assert(relu);
|
||||
return relu;
|
||||
}
|
||||
|
||||
ILayer* reshapeSoftmax(INetworkDefinition *network, ITensor& input, int c) {
|
||||
auto re1 = network->addShuffle(input);
|
||||
assert(re1);
|
||||
re1->setReshapeDimensions(Dims3(c / 2, -1, 0));
|
||||
|
||||
auto sm = network->addSoftMax(*re1->getOutput(0));
|
||||
assert(sm);
|
||||
|
||||
auto re2 = network->addShuffle(*sm->getOutput(0));
|
||||
assert(re2);
|
||||
re2->setReshapeDimensions(Dims3(c, -1, 0));
|
||||
|
||||
return re2;
|
||||
}
|
||||
|
||||
// Creat the engine using only the API and not any parser.
|
||||
ICudaEngine* createEngine(unsigned int maxBatchSize, IBuilder* builder, IBuilderConfig* config, DataType dt) {
|
||||
INetworkDefinition* network = builder->createNetworkV2(0U);
|
||||
|
||||
// Create input tensor of shape {3, INPUT_H, INPUT_W} with name INPUT_BLOB_NAME
|
||||
ITensor* data = network->addInput(INPUT_BLOB_NAME, dt, Dims3{3, INPUT_H, INPUT_W});
|
||||
assert(data);
|
||||
|
||||
std::map<std::string, Weights> weightMap = loadWeights("../retinafaceAntiCov.wts");
|
||||
Weights emptywts{DataType::kFLOAT, nullptr, 0};
|
||||
|
||||
auto conv1 = convBnRelu(network, weightMap, *data, 16, 3, 2, 1, 1, "conv_1");
|
||||
auto conv2 = convBnRelu(network, weightMap, *conv1->getOutput(0), 32, 1, 1, 0, 1, "conv_2");
|
||||
auto conv3dw = convBnRelu(network, weightMap, *conv2->getOutput(0), 32, 3, 2, 1, 32, "conv_3_dw");
|
||||
auto conv3 = convBnRelu(network, weightMap, *conv3dw->getOutput(0), 32, 1, 1, 0, 1, "conv_3");
|
||||
auto conv4dw = convBnRelu(network, weightMap, *conv3->getOutput(0), 32, 3, 1, 1, 32, "conv_4_dw");
|
||||
auto conv4 = convBnRelu(network, weightMap, *conv4dw->getOutput(0), 32, 1, 1, 0, 1, "conv_4");
|
||||
auto conv5dw = convBnRelu(network, weightMap, *conv4->getOutput(0), 32, 3, 2, 1, 32, "conv_5_dw");
|
||||
auto conv5 = convBnRelu(network, weightMap, *conv5dw->getOutput(0), 64, 1, 1, 0, 1, "conv_5");
|
||||
auto conv6dw = convBnRelu(network, weightMap, *conv5->getOutput(0), 64, 3, 1, 1, 64, "conv_6_dw");
|
||||
auto conv6 = convBnRelu(network, weightMap, *conv6dw->getOutput(0), 64, 1, 1, 0, 1, "conv_6");
|
||||
// conv6 to c1
|
||||
auto conv7dw = convBnRelu(network, weightMap, *conv6->getOutput(0), 64, 3, 2, 1, 64, "conv_7_dw");
|
||||
auto conv7 = convBnRelu(network, weightMap, *conv7dw->getOutput(0), 128, 1, 1, 0, 1, "conv_7");
|
||||
auto conv8dw = convBnRelu(network, weightMap, *conv7->getOutput(0), 128, 3, 1, 1, 128, "conv_8_dw");
|
||||
auto conv8 = convBnRelu(network, weightMap, *conv8dw->getOutput(0), 128, 1, 1, 0, 1, "conv_8");
|
||||
auto conv9dw = convBnRelu(network, weightMap, *conv8->getOutput(0), 128, 3, 1, 1, 128, "conv_9_dw");
|
||||
auto conv9 = convBnRelu(network, weightMap, *conv9dw->getOutput(0), 128, 1, 1, 0, 1, "conv_9");
|
||||
auto conv10dw = convBnRelu(network, weightMap, *conv9->getOutput(0), 128, 3, 1, 1, 128, "conv_10_dw");
|
||||
auto conv10 = convBnRelu(network, weightMap, *conv10dw->getOutput(0), 128, 1, 1, 0, 1, "conv_10");
|
||||
auto conv11dw = convBnRelu(network, weightMap, *conv10->getOutput(0), 128, 3, 1, 1, 128, "conv_11_dw");
|
||||
auto conv11 = convBnRelu(network, weightMap, *conv11dw->getOutput(0), 128, 1, 1, 0, 1, "conv_11");
|
||||
auto conv12dw = convBnRelu(network, weightMap, *conv11->getOutput(0), 128, 3, 1, 1, 128, "conv_12_dw");
|
||||
auto conv12 = convBnRelu(network, weightMap, *conv12dw->getOutput(0), 128, 1, 1, 0, 1, "conv_12");
|
||||
// conv12 to c2
|
||||
auto conv13dw = convBnRelu(network, weightMap, *conv12->getOutput(0), 128, 3, 2, 1, 128, "conv_13_dw");
|
||||
auto conv13 = convBnRelu(network, weightMap, *conv13dw->getOutput(0), 256, 1, 1, 0, 1, "conv_13");
|
||||
auto conv14dw = convBnRelu(network, weightMap, *conv13->getOutput(0), 256, 3, 1, 1, 256, "conv_14_dw");
|
||||
auto conv14 = convBnRelu(network, weightMap, *conv14dw->getOutput(0), 256, 1, 1, 0, 1, "conv_14");
|
||||
auto conv_final = convBnRelu(network, weightMap, *conv14->getOutput(0), 256, 1, 1, 0, 1, "conv_final");
|
||||
// convfinal to c3
|
||||
|
||||
auto rf_c3_lateral = convBiasBnRelu(network, weightMap, *conv_final->getOutput(0), 64, 1, 1, 0, "rf_c3_lateral");
|
||||
auto rf_head_s32 = head(network, weightMap, *rf_c3_lateral->getOutput(0), "rf_head_stride32");
|
||||
ILayer *cls_score_s32 = network->addConvolutionNd(*rf_head_s32->getOutput(0), 4, DimsHW{1, 1}, weightMap["face_rpn_cls_score_stride32_weight"], weightMap["face_rpn_cls_score_stride32_bias"]);
|
||||
cls_score_s32 = reshapeSoftmax(network, *cls_score_s32->getOutput(0), 4);
|
||||
auto bbox_s32 = network->addConvolutionNd(*rf_head_s32->getOutput(0), 8, DimsHW{1, 1}, weightMap["face_rpn_bbox_pred_stride32_weight"], weightMap["face_rpn_bbox_pred_stride32_bias"]);
|
||||
auto landmark_s32 = network->addConvolutionNd(*rf_head_s32->getOutput(0), 20, DimsHW{1, 1}, weightMap["face_rpn_landmark_pred_stride32_weight"], weightMap["face_rpn_landmark_pred_stride32_bias"]);
|
||||
auto rf_head2_s32 = head(network, weightMap, *rf_c3_lateral->getOutput(0), "rf_head2_stride32");
|
||||
ILayer *type_score_s32 = network->addConvolutionNd(*rf_head2_s32->getOutput(0), 6, DimsHW{1, 1}, weightMap["face_rpn_type_score_stride32_weight"], weightMap["face_rpn_type_score_stride32_bias"]);
|
||||
type_score_s32 = reshapeSoftmax(network, *type_score_s32->getOutput(0), 6);
|
||||
|
||||
float *deval = reinterpret_cast<float*>(malloc(sizeof(float) * 64 * 2 * 2));
|
||||
for (int i = 0; i < 64 * 2 * 2; i++) {
|
||||
deval[i] = 1.0;
|
||||
}
|
||||
Weights deconvwts{DataType::kFLOAT, deval, 64 * 2 * 2};
|
||||
IDeconvolutionLayer* c3_deconv = network->addDeconvolutionNd(*rf_c3_lateral->getOutput(0), 64, DimsHW{2, 2}, deconvwts, emptywts);
|
||||
assert(c3_deconv);
|
||||
c3_deconv->setStrideNd(DimsHW{2, 2});
|
||||
c3_deconv->setNbGroups(64);
|
||||
weightMap["c3_deconv"] = deconvwts;
|
||||
auto rf_c2_lateral = convBiasBnRelu(network, weightMap, *conv12->getOutput(0), 64, 1, 1, 0, "rf_c2_lateral");
|
||||
auto plus0 = network->addElementWise(*c3_deconv->getOutput(0), *rf_c2_lateral->getOutput(0), ElementWiseOperation::kSUM);
|
||||
auto rf_c2_aggr = convBiasBnRelu(network, weightMap, *plus0->getOutput(0), 64, 3, 1, 1, "rf_c2_aggr");
|
||||
auto rf_head_s16 = head(network, weightMap, *rf_c2_aggr->getOutput(0), "rf_head_stride16");
|
||||
ILayer *cls_score_s16 = network->addConvolutionNd(*rf_head_s16->getOutput(0), 4, DimsHW{1, 1}, weightMap["face_rpn_cls_score_stride16_weight"], weightMap["face_rpn_cls_score_stride16_bias"]);
|
||||
cls_score_s16 = reshapeSoftmax(network, *cls_score_s16->getOutput(0), 4);
|
||||
auto bbox_s16 = network->addConvolutionNd(*rf_head_s16->getOutput(0), 8, DimsHW{1, 1}, weightMap["face_rpn_bbox_pred_stride16_weight"], weightMap["face_rpn_bbox_pred_stride16_bias"]);
|
||||
auto landmark_s16 = network->addConvolutionNd(*rf_head_s16->getOutput(0), 20, DimsHW{1, 1}, weightMap["face_rpn_landmark_pred_stride16_weight"], weightMap["face_rpn_landmark_pred_stride16_bias"]);
|
||||
auto rf_head2_s16 = head(network, weightMap, *rf_c2_aggr->getOutput(0), "rf_head2_stride16");
|
||||
ILayer *type_score_s16 = network->addConvolutionNd(*rf_head2_s16->getOutput(0), 6, DimsHW{1, 1}, weightMap["face_rpn_type_score_stride16_weight"], weightMap["face_rpn_type_score_stride16_bias"]);
|
||||
type_score_s16 = reshapeSoftmax(network, *type_score_s16->getOutput(0), 6);
|
||||
|
||||
IDeconvolutionLayer* c2_deconv = network->addDeconvolutionNd(*rf_c2_aggr->getOutput(0), 64, DimsHW{2, 2}, deconvwts, emptywts);
|
||||
assert(c2_deconv);
|
||||
c2_deconv->setStrideNd(DimsHW{2, 2});
|
||||
c2_deconv->setNbGroups(64);
|
||||
auto rf_c1_red = convBiasBnRelu(network, weightMap, *conv6->getOutput(0), 64, 1, 1, 0, "rf_c1_red_conv");
|
||||
auto plus1 = network->addElementWise(*c2_deconv->getOutput(0), *rf_c1_red->getOutput(0), ElementWiseOperation::kSUM);
|
||||
auto rf_c1_aggr = convBiasBnRelu(network, weightMap, *plus1->getOutput(0), 64, 3, 1, 1, "rf_c1_aggr");
|
||||
auto rf_head_s8 = head(network, weightMap, *rf_c1_aggr->getOutput(0), "rf_head_stride8");
|
||||
ILayer *cls_score_s8 = network->addConvolutionNd(*rf_head_s8->getOutput(0), 4, DimsHW{1, 1}, weightMap["face_rpn_cls_score_stride8_weight"], weightMap["face_rpn_cls_score_stride8_bias"]);
|
||||
cls_score_s8 = reshapeSoftmax(network, *cls_score_s8->getOutput(0), 4);
|
||||
auto bbox_s8 = network->addConvolutionNd(*rf_head_s8->getOutput(0), 8, DimsHW{1, 1}, weightMap["face_rpn_bbox_pred_stride8_weight"], weightMap["face_rpn_bbox_pred_stride8_bias"]);
|
||||
auto landmark_s8 = network->addConvolutionNd(*rf_head_s8->getOutput(0), 20, DimsHW{1, 1}, weightMap["face_rpn_landmark_pred_stride8_weight"], weightMap["face_rpn_landmark_pred_stride8_bias"]);
|
||||
auto rf_head2_s8 = head(network, weightMap, *rf_c1_aggr->getOutput(0), "rf_head2_stride8");
|
||||
ILayer *type_score_s8 = network->addConvolutionNd(*rf_head2_s8->getOutput(0), 6, DimsHW{1, 1}, weightMap["face_rpn_type_score_stride8_weight"], weightMap["face_rpn_type_score_stride8_bias"]);
|
||||
type_score_s8 = reshapeSoftmax(network, *type_score_s8->getOutput(0), 6);
|
||||
|
||||
ITensor* inputTensors_s32[] = {cls_score_s32->getOutput(0), bbox_s32->getOutput(0), landmark_s32->getOutput(0), type_score_s32->getOutput(0)};
|
||||
auto cat_s32 = network->addConcatenation(inputTensors_s32, 4);
|
||||
assert(cat_s32);
|
||||
|
||||
ITensor* inputTensors_s16[] = {cls_score_s16->getOutput(0), bbox_s16->getOutput(0), landmark_s16->getOutput(0), type_score_s16->getOutput(0)};
|
||||
auto cat_s16 = network->addConcatenation(inputTensors_s16, 4);
|
||||
assert(cat_s16);
|
||||
|
||||
ITensor* inputTensors_s8[] = {cls_score_s8->getOutput(0), bbox_s8->getOutput(0), landmark_s8->getOutput(0), type_score_s8->getOutput(0)};
|
||||
auto cat_s8 = network->addConcatenation(inputTensors_s8, 4);
|
||||
assert(cat_s8);
|
||||
|
||||
auto creator = getPluginRegistry()->getPluginCreator("Decode_TRT", "1");
|
||||
PluginFieldCollection pfc;
|
||||
IPluginV2 *pluginObj = creator->createPlugin("decode", &pfc);
|
||||
ITensor* inputTensors[] = {cat_s8->getOutput(0), cat_s16->getOutput(0), cat_s32->getOutput(0)};
|
||||
auto decodelayer = network->addPluginV2(inputTensors, 3, *pluginObj);
|
||||
assert(decodelayer);
|
||||
|
||||
decodelayer->getOutput(0)->setName(OUTPUT_BLOB_NAME);
|
||||
network->markOutput(*decodelayer->getOutput(0));
|
||||
|
||||
// Build engine
|
||||
builder->setMaxBatchSize(maxBatchSize);
|
||||
config->setMaxWorkspaceSize(16 * (1 << 20)); // 16MB
|
||||
#ifdef USE_FP16
|
||||
config->setFlag(BuilderFlag::kFP16);
|
||||
#endif
|
||||
std::cout << "Building engine, please wait for a while..." << std::endl;
|
||||
ICudaEngine* engine = builder->buildEngineWithConfig(*network, *config);
|
||||
std::cout << "Build engine successfully!" << std::endl;
|
||||
|
||||
// Don't need the network any more
|
||||
network->destroy();
|
||||
|
||||
// Release host memory
|
||||
for (auto& mem : weightMap)
|
||||
{
|
||||
free((void*) (mem.second.values));
|
||||
}
|
||||
|
||||
return engine;
|
||||
}
|
||||
|
||||
void APIToModel(unsigned int maxBatchSize, IHostMemory** modelStream) {
|
||||
// Create builder
|
||||
IBuilder* builder = createInferBuilder(gLogger);
|
||||
IBuilderConfig* config = builder->createBuilderConfig();
|
||||
|
||||
// Create model to populate the network, then set the outputs and create an engine
|
||||
ICudaEngine* engine = createEngine(maxBatchSize, builder, config, DataType::kFLOAT);
|
||||
assert(engine != nullptr);
|
||||
|
||||
// Serialize the engine
|
||||
(*modelStream) = engine->serialize();
|
||||
|
||||
// Close everything down
|
||||
engine->destroy();
|
||||
builder->destroy();
|
||||
}
|
||||
|
||||
void doInference(IExecutionContext& context, float* input, float* output, int batchSize) {
|
||||
const ICudaEngine& engine = context.getEngine();
|
||||
|
||||
// Pointers to input and output device buffers to pass to engine.
|
||||
// Engine requires exactly IEngine::getNbBindings() number of buffers.
|
||||
assert(engine.getNbBindings() == 2);
|
||||
void* buffers[2];
|
||||
|
||||
// In order to bind the buffers, we need to know the names of the input and output tensors.
|
||||
// Note that indices are guaranteed to be less than IEngine::getNbBindings()
|
||||
const int inputIndex = engine.getBindingIndex(INPUT_BLOB_NAME);
|
||||
const int outputIndex = engine.getBindingIndex(OUTPUT_BLOB_NAME);
|
||||
|
||||
// Create GPU buffers on device
|
||||
CHECK(cudaMalloc(&buffers[inputIndex], batchSize * 3 * INPUT_H * INPUT_W * sizeof(float)));
|
||||
CHECK(cudaMalloc(&buffers[outputIndex], batchSize * OUTPUT_SIZE * sizeof(float)));
|
||||
|
||||
// Create stream
|
||||
cudaStream_t stream;
|
||||
CHECK(cudaStreamCreate(&stream));
|
||||
|
||||
// DMA input batch data to device, infer on the batch asynchronously, and DMA output back to host
|
||||
CHECK(cudaMemcpyAsync(buffers[inputIndex], input, batchSize * 3 * INPUT_H * INPUT_W * sizeof(float), cudaMemcpyHostToDevice, stream));
|
||||
context.enqueue(batchSize, buffers, stream, nullptr);
|
||||
CHECK(cudaMemcpyAsync(output, buffers[outputIndex], batchSize * OUTPUT_SIZE * sizeof(float), cudaMemcpyDeviceToHost, stream));
|
||||
cudaStreamSynchronize(stream);
|
||||
|
||||
// Release stream and buffers
|
||||
cudaStreamDestroy(stream);
|
||||
CHECK(cudaFree(buffers[inputIndex]));
|
||||
CHECK(cudaFree(buffers[outputIndex]));
|
||||
}
|
||||
|
||||
int read_files_in_dir(const char *p_dir_name, std::vector<std::string> &file_names) {
|
||||
DIR *p_dir = opendir(p_dir_name);
|
||||
if (p_dir == nullptr) {
|
||||
return -1;
|
||||
}
|
||||
|
||||
struct dirent* p_file = nullptr;
|
||||
while ((p_file = readdir(p_dir)) != nullptr) {
|
||||
if (strcmp(p_file->d_name, ".") != 0 &&
|
||||
strcmp(p_file->d_name, "..") != 0) {
|
||||
//std::string cur_file_name(p_dir_name);
|
||||
//cur_file_name += "/";
|
||||
//cur_file_name += p_file->d_name;
|
||||
std::string cur_file_name(p_file->d_name);
|
||||
file_names.push_back(cur_file_name);
|
||||
}
|
||||
}
|
||||
|
||||
closedir(p_dir);
|
||||
return 0;
|
||||
}
|
||||
|
||||
int main(int argc, char** argv) {
|
||||
cudaSetDevice(DEVICE);
|
||||
// create a model using the API directly and serialize it to a stream
|
||||
char *trtModelStream{nullptr};
|
||||
size_t size{0};
|
||||
|
||||
if (argc == 2 && std::string(argv[1]) == "-s") {
|
||||
IHostMemory* modelStream{nullptr};
|
||||
APIToModel(BATCH_SIZE, &modelStream);
|
||||
assert(modelStream != nullptr);
|
||||
std::ofstream p("retinafaceAntiCov.engine", std::ios::binary);
|
||||
if (!p) {
|
||||
std::cerr << "could not open plan output file" << std::endl;
|
||||
return -1;
|
||||
}
|
||||
p.write(reinterpret_cast<const char*>(modelStream->data()), modelStream->size());
|
||||
modelStream->destroy();
|
||||
return 0;
|
||||
} else if (argc == 2 && std::string(argv[1]) == "-d") {
|
||||
std::ifstream file("retinafaceAntiCov.engine", std::ios::binary);
|
||||
if (file.good()) {
|
||||
file.seekg(0, file.end);
|
||||
size = file.tellg();
|
||||
file.seekg(0, file.beg);
|
||||
trtModelStream = new char[size];
|
||||
assert(trtModelStream);
|
||||
file.read(trtModelStream, size);
|
||||
file.close();
|
||||
}
|
||||
} else {
|
||||
std::cerr << "arguments not right!" << std::endl;
|
||||
std::cerr << "./retinafaceAntiCov -s // serialize model to plan file" << std::endl;
|
||||
std::cerr << "./retinafaceAntiCov -d // deserialize plan file and run inference" << std::endl;
|
||||
return -1;
|
||||
}
|
||||
|
||||
// prepare input data ---------------------------
|
||||
static float data[BATCH_SIZE * 3 * INPUT_H * INPUT_W];
|
||||
//for (int i = 0; i < 3 * INPUT_H * INPUT_W; i++)
|
||||
// data[i] = 1.0;
|
||||
static float prob[BATCH_SIZE * OUTPUT_SIZE];
|
||||
IRuntime* runtime = createInferRuntime(gLogger);
|
||||
assert(runtime != nullptr);
|
||||
ICudaEngine* engine = runtime->deserializeCudaEngine(trtModelStream, size);
|
||||
assert(engine != nullptr);
|
||||
IExecutionContext* context = engine->createExecutionContext();
|
||||
assert(context != nullptr);
|
||||
delete[] trtModelStream;
|
||||
|
||||
cv::Mat img = cv::imread("test.jpg");
|
||||
cv::Mat pr_img = preprocess_img(img);
|
||||
for (int i = 0; i < INPUT_H * INPUT_W; i++) {
|
||||
data[i] = ((float)pr_img.at<cv::Vec3b>(i)[2] - 127.5) * 0.0078125;
|
||||
data[i + INPUT_H * INPUT_W] = ((float)pr_img.at<cv::Vec3b>(i)[1] - 127.5) * 0.0078125;
|
||||
data[i + 2 * INPUT_H * INPUT_W] = ((float)pr_img.at<cv::Vec3b>(i)[0] - 127.5) * 0.0078125;
|
||||
}
|
||||
|
||||
// Run inference
|
||||
auto start = std::chrono::system_clock::now();
|
||||
doInference(*context, data, prob, BATCH_SIZE);
|
||||
auto end = std::chrono::system_clock::now();
|
||||
std::cout << std::chrono::duration_cast<std::chrono::milliseconds>(end - start).count() << "ms" << std::endl;
|
||||
|
||||
std::vector<decodeplugin::Detection> res;
|
||||
nms(res, prob);
|
||||
|
||||
for (size_t j = 0; j < res.size(); j++) {
|
||||
//if (res[j].class_confidence < 0.1) continue;
|
||||
cv::Rect r = get_rect_adapt_landmark(img, res[j].bbox, res[j].landmark);
|
||||
cv::rectangle(img, r, cv::Scalar(0x27, 0xC1, 0x36), 2);
|
||||
cv::putText(img, "face: " + std::to_string((int)(res[j].class_confidence * 100)) + "%", cv::Point(r.x, r.y + 20), cv::FONT_HERSHEY_PLAIN, 1.2, cv::Scalar(0xFF, 0xFF, 0xFF), 1);
|
||||
for (int k = 0; k < 10; k += 2) {
|
||||
cv::circle(img, cv::Point(res[j].landmark[k], res[j].landmark[k + 1]), 1, cv::Scalar(255 * (k > 2), 255 * (k > 0 && k < 8), 255 * (k < 6)), 4);
|
||||
}
|
||||
cv::putText(img, "mask: " + std::to_string((int)(res[j].mask_confidence * 100)) + "%", cv::Point(r.x, r.y + 40), cv::FONT_HERSHEY_PLAIN, 1.2, cv::Scalar(0x00, 0x00, 0xFF), 1);
|
||||
}
|
||||
cv::imwrite("out.jpg", img);
|
||||
|
||||
// Destroy the engine
|
||||
context->destroy();
|
||||
engine->destroy();
|
||||
runtime->destroy();
|
||||
|
||||
//Print histogram of the output distribution
|
||||
//std::cout << "\nOutput:\n\n";
|
||||
//for (unsigned int i = 0; i < OUTPUT_SIZE; i++)
|
||||
//{
|
||||
// std::cout << prob[i] << ", ";
|
||||
// if (i % 10 == 0) std::cout << i / 10 << std::endl;
|
||||
//}
|
||||
//std::cout << std::endl;
|
||||
|
||||
return 0;
|
||||
}
|
||||
Loading…
Reference in New Issue
Block a user