From c4fd9ca69753bc8b4fab299d37a795ebe4e60f8e Mon Sep 17 00:00:00 2001 From: wang-xinyu Date: Sun, 24 Jan 2021 14:12:50 +0800 Subject: [PATCH] update readme --- README.md | 4 +--- arcface/README.md | 20 ++++++++++++-------- yolov5/README.md | 2 +- 3 files changed, 14 insertions(+), 12 deletions(-) diff --git a/README.md b/README.md index bf7b818..ca6a79c 100644 --- a/README.md +++ b/README.md @@ -10,6 +10,7 @@ All the models are implemented in pytorch/mxnet/tensorflown first, and export a ## News +- `24 Jan 2021`. IBN-Net added by [TCHeish](https://github.com/TCHeish), PSENet optimized, YOLOv5 v4.0 INT8, etc. - `8 Jan 2021`. YOLOv5 s/m/l/x updated to v4.0. - `27 Dec 2020`. HRNet-Semantic-Segmentation added by [BaofengZan](https://github.com/BaofengZan). - `4 Dec 2020`. DBNet dynamic input shape support by [BaofengZan](https://github.com/BaofengZan), YOLOv3 int8, PSENet(tensorflow) text detection by [upczww](https://github.com/upczww). @@ -18,13 +19,10 @@ All the models are implemented in pytorch/mxnet/tensorflown first, and export a - `7 Nov 2020`. All models migrated to trt7 API, and clean up the master branch. - `29 Oct 2020`. First INT8 quantization implementation! Please check retinaface. - `23 Oct 2020`. Add a .wts model zoo for quick evaluation. -- `21 Sep 2020`. [BaofengZan](https://github.com/BaofengZan) added hrnet classification and step by step tutorial(Chinese). - `16 Sep 2020`. [hwh-hit](https://github.com/hwh-hit) added ufld(Ultra-Fast-Lane-Detection, ECCV2020). - `13 Sep 2020`. Add crnn, and got 1000fps on GTX1080. - `28 Aug 2020`. [BaofengZan](https://github.com/BaofengZan) added a tutorial for compiling and running tensorrtx on windows. - `16 Aug 2020`. [upczww](https://github.com/upczww) added a python wrapper for yolov5. -- `14 Aug 2020`. Update yolov5 to v3.0 release. -- `3 Aug 2020`. [BaofengZan](https://github.com/BaofengZan) implemented yolov5 s/m/l/x (yolov5 v2.0 release). - `28 May 2020`. arcface LResNet50E-IR model from [deepinsight/insightface](https://github.com/deepinsight/insightface) implemented. We got 333fps on GTX1080. - `22 May 2020`. A new branch [trt4](https://github.com/wang-xinyu/tensorrtx/tree/trt4) created, which is using TensorRT 4 API. Now the master branch is using TensorRT 7 API. But only `yolov4` has been migrated to TensorRT 7 API for now. The rest will be migrated soon. And a tutorial for `migarating from TensorRT 4 to 7` provided. diff --git a/arcface/README.md b/arcface/README.md index c0df3ec..11b46a4 100644 --- a/arcface/README.md +++ b/arcface/README.md @@ -10,21 +10,30 @@ The two input images used in this project are joey0.ppm and joey1.ppm, download

+## Config + +- FP16/FP32 can be selected by the macro `USE_FP16` in arcface-r50.cpp +- GPU id can be selected by the macro `DEVICE` in arcface-r50.cpp + ## Run -``` 1. generate arcface-r50.wts from mxnet implementation with LResNet50E-IR,ArcFace@ms1m-refine-v1 pretrained model +``` git clone https://github.com/deepinsight/insightface -cd insightface/deploy +cd insightface +git checkout 3866cd77a6896c934b51ed39e9651b791d78bb57 +cd deploy // copy tensorrtx/arcface/gen_wts.py to here(insightface/deploy) // download model-r50-am-lfw.zip and unzip here(insightface/deploy) python gen_wts.py // a file 'arcface-r50.wts' will be generated. // the master branch of insightface should work, if not, you can checkout 94ad870abb3203d6f31b049b70dd080dc8f33fca +``` 2. put arcface-r50.wts into tensorrtx/arcface, build and run +``` cd tensorrtx/arcface // download joey0.ppm and joey1.ppm, and put here(tensorrtx/arcface) mkdir build @@ -33,14 +42,9 @@ cmake .. make sudo ./arcface-r50 -s // serialize model to plan file i.e. 'arcface-r50.engine' sudo ./arcface-r50 -d // deserialize plan file and run inference - -3. check the output log, latency and similarity score. ``` -## Config - -- FP16/FP32 can be selected by the macro `USE_FP16` in arcface-r50.cpp -- GPU id can be selected by the macro `DEVICE` in arcface-r50.cpp +3. check the output log, latency and similarity score. ## More Information diff --git a/yolov5/README.md b/yolov5/README.md index a2eda5a..dc6175f 100644 --- a/yolov5/README.md +++ b/yolov5/README.md @@ -66,7 +66,7 @@ python yolov5_trt.py 2. unzip it in yolov5/build -3. set the macro `USE_INT8` in yolov3.cpp and make +3. set the macro `USE_INT8` in yolov5.cpp and make 4. serialize the model and test