haotian/duan8
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

yolov5 updated to v4.0

Browse Source
This commit is contained in:
wang-xinyu 2021-01-08 16:52:48 +08:00
parent 6d0f5cbf47
commit d0a32e5646
3 changed files with 56 additions and 41 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

6
yolov5/README.md
View File

@ -2,10 +2,10 @@
The Pytorch implementation is [ultralytics/yolov5](https://github.com/ultralytics/yolov5).
Currently, we support yolov5 v1.0(yolov5s only), v2.0, v3.0 and v3.1.
Currently, we support yolov5 v1.0(yolov5s only), v2.0, v3.0, v3.1 and v4.0.
- For yolov5 v3.1, please visit [yolov5 release v3.1](https://github.com/ultralytics/yolov5/releases/tag/v3.1), and use the latest commit of this repo.
- For yolov5 v3.0, please visit [yolov5 release v3.0](https://github.com/ultralytics/yolov5/releases/tag/v3.0), and use the latest commit of this repo.
- For yolov5 v4.0, please visit [yolov5 release v4.0](https://github.com/ultralytics/yolov5/releases/tag/v4.0), and use the latest commit of this repo.
- For yolov5 v3.0 and v3.1, please visit [yolov5 release v3.0](https://github.com/ultralytics/yolov5/releases/tag/v3.0) and [yolov5 release v3.1](https://github.com/ultralytics/yolov5/releases/tag/v3.1), and checkout commit ['6d0f5cb'](https://github.com/wang-xinyu/tensorrtx/commit/6d0f5cbf4745bc00b69aad54a905383fb906f103) of this repo.
- For yolov5 v2.0, please visit [yolov5 release v2.0](https://github.com/ultralytics/yolov5/releases/tag/v2.0), and checkout commit ['5cfa444'](https://github.com/wang-xinyu/tensorrtx/commit/5cfa4445170eabaa54acd5ad7f469ef65a8763f1) of this repo.
- For yolov5 v1.0, please visit [yolov5 release v1.0](https://github.com/ultralytics/yolov5/releases/tag/v1.0), and checkout commit ['f09aa3b'](https://github.com/wang-xinyu/tensorrtx/commit/f09aa3bbebf4d4d37b6d3b32a1d39e1f2678a07b) of this repo.

27
yolov5/common.hpp
View File

@ -200,12 +200,10 @@ ILayer* convBlock(INetworkDefinition *network, std::map<std::string, Weights>& w
conv1->setNbGroups(g);
IScaleLayer* bn1 = addBatchNorm2d(network, weightMap, *conv1->getOutput(0), lname + ".bn", 1e-3);
// hard_swish = x * hard_sigmoid
auto hsig = network->addActivation(*bn1->getOutput(0), ActivationType::kHARD_SIGMOID);
assert(hsig);
hsig->setAlpha(1.0 / 6.0);
hsig->setBeta(0.5);
auto ew = network->addElementWise(*bn1->getOutput(0), *hsig->getOutput(0), ElementWiseOperation::kPROD);
// silu = x * sigmoid
auto sig = network->addActivation(*bn1->getOutput(0), ActivationType::kSIGMOID);
assert(sig);
auto ew = network->addElementWise(*bn1->getOutput(0), *sig->getOutput(0), ElementWiseOperation::kPROD);
assert(ew);
return ew;
}
@ -254,6 +252,23 @@ ILayer* bottleneckCSP(INetworkDefinition *network, std::map<std::string, Weights
return cv4;
}
ILayer* C3(INetworkDefinition *network, std::map<std::string, Weights>& weightMap, ITensor& input, int c1, int c2, int n, bool shortcut, int g, float e, std::string lname) {
int c_ = (int)((float)c2 * e);
auto cv1 = convBlock(network, weightMap, input, c_, 1, 1, 1, lname + ".cv1");
auto cv2 = convBlock(network, weightMap, input, c_, 1, 1, 1, lname + ".cv2");
ITensor *y1 = cv1->getOutput(0);
for (int i = 0; i < n; i++) {
auto b = bottleneck(network, weightMap, *y1, c_, c_, shortcut, g, 1.0, lname + ".m." + std::to_string(i));
y1 = b->getOutput(0);
}
ITensor* inputTensors[] = { y1, cv2->getOutput(0) };
auto cat = network->addConcatenation(inputTensors, 2);
auto cv3 = convBlock(network, weightMap, *cat->getOutput(0), c2, 1, 1, 1, lname + ".cv3");
return cv3;
}
ILayer* SPP(INetworkDefinition *network, std::map<std::string, Weights>& weightMap, ITensor& input, int c1, int c2, int k1, int k2, int k3, std::string lname) {
int c_ = c1 / 2;
auto cv1 = convBlock(network, weightMap, input, c_, 1, 1, 1, lname + ".cv1");

64
yolov5/yolov5.cpp
View File

@ -39,16 +39,16 @@ ICudaEngine* createEngine_s(unsigned int maxBatchSize, IBuilder* builder, IBuild
// yolov5 backbone
auto focus0 = focus(network, weightMap, *data, 3, 32, 3, "model.0");
auto conv1 = convBlock(network, weightMap, *focus0->getOutput(0), 64, 3, 2, 1, "model.1");
auto bottleneck_CSP2 = bottleneckCSP(network, weightMap, *conv1->getOutput(0), 64, 64, 1, true, 1, 0.5, "model.2");
auto bottleneck_CSP2 = C3(network, weightMap, *conv1->getOutput(0), 64, 64, 1, true, 1, 0.5, "model.2");
auto conv3 = convBlock(network, weightMap, *bottleneck_CSP2->getOutput(0), 128, 3, 2, 1, "model.3");
auto bottleneck_csp4 = bottleneckCSP(network, weightMap, *conv3->getOutput(0), 128, 128, 3, true, 1, 0.5, "model.4");
auto bottleneck_csp4 = C3(network, weightMap, *conv3->getOutput(0), 128, 128, 3, true, 1, 0.5, "model.4");
auto conv5 = convBlock(network, weightMap, *bottleneck_csp4->getOutput(0), 256, 3, 2, 1, "model.5");
auto bottleneck_csp6 = bottleneckCSP(network, weightMap, *conv5->getOutput(0), 256, 256, 3, true, 1, 0.5, "model.6");
auto bottleneck_csp6 = C3(network, weightMap, *conv5->getOutput(0), 256, 256, 3, true, 1, 0.5, "model.6");
auto conv7 = convBlock(network, weightMap, *bottleneck_csp6->getOutput(0), 512, 3, 2, 1, "model.7");
auto spp8 = SPP(network, weightMap, *conv7->getOutput(0), 512, 512, 5, 9, 13, "model.8");
// yolov5 head
auto bottleneck_csp9 = bottleneckCSP(network, weightMap, *spp8->getOutput(0), 512, 512, 1, false, 1, 0.5, "model.9");
auto bottleneck_csp9 = C3(network, weightMap, *spp8->getOutput(0), 512, 512, 1, false, 1, 0.5, "model.9");
auto conv10 = convBlock(network, weightMap, *bottleneck_csp9->getOutput(0), 256, 1, 1, 1, "model.10");
float *deval = reinterpret_cast<float*>(malloc(sizeof(float) * 256 * 2 * 2));
@ -63,7 +63,7 @@ ICudaEngine* createEngine_s(unsigned int maxBatchSize, IBuilder* builder, IBuild
ITensor* inputTensors12[] = { deconv11->getOutput(0), bottleneck_csp6->getOutput(0) };
auto cat12 = network->addConcatenation(inputTensors12, 2);
auto bottleneck_csp13 = bottleneckCSP(network, weightMap, *cat12->getOutput(0), 512, 256, 1, false, 1, 0.5, "model.13");
auto bottleneck_csp13 = C3(network, weightMap, *cat12->getOutput(0), 512, 256, 1, false, 1, 0.5, "model.13");
auto conv14 = convBlock(network, weightMap, *bottleneck_csp13->getOutput(0), 128, 1, 1, 1, "model.14");
Weights deconvwts15{ DataType::kFLOAT, deval, 128 * 2 * 2 };
@ -73,19 +73,19 @@ ICudaEngine* createEngine_s(unsigned int maxBatchSize, IBuilder* builder, IBuild
ITensor* inputTensors16[] = { deconv15->getOutput(0), bottleneck_csp4->getOutput(0) };
auto cat16 = network->addConcatenation(inputTensors16, 2);
auto bottleneck_csp17 = bottleneckCSP(network, weightMap, *cat16->getOutput(0), 256, 128, 1, false, 1, 0.5, "model.17");
auto bottleneck_csp17 = C3(network, weightMap, *cat16->getOutput(0), 256, 128, 1, false, 1, 0.5, "model.17");
IConvolutionLayer* det0 = network->addConvolutionNd(*bottleneck_csp17->getOutput(0), 3 * (Yolo::CLASS_NUM + 5), DimsHW{ 1, 1 }, weightMap["model.24.m.0.weight"], weightMap["model.24.m.0.bias"]);
auto conv18 = convBlock(network, weightMap, *bottleneck_csp17->getOutput(0), 128, 3, 2, 1, "model.18");
ITensor* inputTensors19[] = { conv18->getOutput(0), conv14->getOutput(0) };
auto cat19 = network->addConcatenation(inputTensors19, 2);
auto bottleneck_csp20 = bottleneckCSP(network, weightMap, *cat19->getOutput(0), 256, 256, 1, false, 1, 0.5, "model.20");
auto bottleneck_csp20 = C3(network, weightMap, *cat19->getOutput(0), 256, 256, 1, false, 1, 0.5, "model.20");
IConvolutionLayer* det1 = network->addConvolutionNd(*bottleneck_csp20->getOutput(0), 3 * (Yolo::CLASS_NUM + 5), DimsHW{ 1, 1 }, weightMap["model.24.m.1.weight"], weightMap["model.24.m.1.bias"]);
auto conv21 = convBlock(network, weightMap, *bottleneck_csp20->getOutput(0), 256, 3, 2, 1, "model.21");
ITensor* inputTensors22[] = { conv21->getOutput(0), conv10->getOutput(0) };
auto cat22 = network->addConcatenation(inputTensors22, 2);
auto bottleneck_csp23 = bottleneckCSP(network, weightMap, *cat22->getOutput(0), 512, 512, 1, false, 1, 0.5, "model.23");
auto bottleneck_csp23 = C3(network, weightMap, *cat22->getOutput(0), 512, 512, 1, false, 1, 0.5, "model.23");
IConvolutionLayer* det2 = network->addConvolutionNd(*bottleneck_csp23->getOutput(0), 3 * (Yolo::CLASS_NUM + 5), DimsHW{ 1, 1 }, weightMap["model.24.m.2.weight"], weightMap["model.24.m.2.bias"]);
auto yolo = addYoLoLayer(network, weightMap, det0, det1, det2);
@ -127,15 +127,15 @@ ICudaEngine* createEngine_m(unsigned int maxBatchSize, IBuilder* builder, IBuild
/* ------ yolov5 backbone------ */
auto focus0 = focus(network, weightMap, *data, 3, 48, 3, "model.0");
auto conv1 = convBlock(network, weightMap, *focus0->getOutput(0), 96, 3, 2, 1, "model.1");
auto bottleneck_CSP2 = bottleneckCSP(network, weightMap, *conv1->getOutput(0), 96, 96, 2, true, 1, 0.5, "model.2");
auto bottleneck_CSP2 = C3(network, weightMap, *conv1->getOutput(0), 96, 96, 2, true, 1, 0.5, "model.2");
auto conv3 = convBlock(network, weightMap, *bottleneck_CSP2->getOutput(0), 192, 3, 2, 1, "model.3");
auto bottleneck_csp4 = bottleneckCSP(network, weightMap, *conv3->getOutput(0), 192, 192, 6, true, 1, 0.5, "model.4");
auto bottleneck_csp4 = C3(network, weightMap, *conv3->getOutput(0), 192, 192, 6, true, 1, 0.5, "model.4");
auto conv5 = convBlock(network, weightMap, *bottleneck_csp4->getOutput(0), 384, 3, 2, 1, "model.5");
auto bottleneck_csp6 = bottleneckCSP(network, weightMap, *conv5->getOutput(0), 384, 384, 6, true, 1, 0.5, "model.6");
auto bottleneck_csp6 = C3(network, weightMap, *conv5->getOutput(0), 384, 384, 6, true, 1, 0.5, "model.6");
auto conv7 = convBlock(network, weightMap, *bottleneck_csp6->getOutput(0), 768, 3, 2, 1, "model.7");
auto spp8 = SPP(network, weightMap, *conv7->getOutput(0), 768, 768, 5, 9, 13, "model.8");
/* ------ yolov5 head ------ */
auto bottleneck_csp9 = bottleneckCSP(network, weightMap, *spp8->getOutput(0), 768, 768, 2, false, 1, 0.5, "model.9");
auto bottleneck_csp9 = C3(network, weightMap, *spp8->getOutput(0), 768, 768, 2, false, 1, 0.5, "model.9");
auto conv10 = convBlock(network, weightMap, *bottleneck_csp9->getOutput(0), 384, 1, 1, 1, "model.10");
float *deval = reinterpret_cast<float*>(malloc(sizeof(float) * 384 * 2 * 2));
@ -150,7 +150,7 @@ ICudaEngine* createEngine_m(unsigned int maxBatchSize, IBuilder* builder, IBuild
ITensor* inputTensors12[] = { deconv11->getOutput(0), bottleneck_csp6->getOutput(0) };
auto cat12 = network->addConcatenation(inputTensors12, 2);
auto bottleneck_csp13 = bottleneckCSP(network, weightMap, *cat12->getOutput(0), 768, 384, 2, false, 1, 0.5, "model.13");
auto bottleneck_csp13 = C3(network, weightMap, *cat12->getOutput(0), 768, 384, 2, false, 1, 0.5, "model.13");
auto conv14 = convBlock(network, weightMap, *bottleneck_csp13->getOutput(0), 192, 1, 1, 1, "model.14");
@ -161,21 +161,21 @@ ICudaEngine* createEngine_m(unsigned int maxBatchSize, IBuilder* builder, IBuild
ITensor* inputTensors16[] = { deconv15->getOutput(0), bottleneck_csp4->getOutput(0) };
auto cat16 = network->addConcatenation(inputTensors16, 2);
auto bottleneck_csp17 = bottleneckCSP(network, weightMap, *cat16->getOutput(0), 384, 192, 2, false, 1, 0.5, "model.17");
auto bottleneck_csp17 = C3(network, weightMap, *cat16->getOutput(0), 384, 192, 2, false, 1, 0.5, "model.17");
//yolo layer 0
IConvolutionLayer* det0 = network->addConvolutionNd(*bottleneck_csp17->getOutput(0), 3 * (Yolo::CLASS_NUM + 5), DimsHW{ 1, 1 }, weightMap["model.24.m.0.weight"], weightMap["model.24.m.0.bias"]);
auto conv18 = convBlock(network, weightMap, *bottleneck_csp17->getOutput(0), 192, 3, 2, 1, "model.18");
ITensor* inputTensors19[] = { conv18->getOutput(0), conv14->getOutput(0) };
auto cat19 = network->addConcatenation(inputTensors19, 2);
auto bottleneck_csp20 = bottleneckCSP(network, weightMap, *cat19->getOutput(0), 384, 384, 2, false, 1, 0.5, "model.20");
auto bottleneck_csp20 = C3(network, weightMap, *cat19->getOutput(0), 384, 384, 2, false, 1, 0.5, "model.20");
//yolo layer 1
IConvolutionLayer* det1 = network->addConvolutionNd(*bottleneck_csp20->getOutput(0), 3 * (Yolo::CLASS_NUM + 5), DimsHW{ 1, 1 }, weightMap["model.24.m.1.weight"], weightMap["model.24.m.1.bias"]);
auto conv21 = convBlock(network, weightMap, *bottleneck_csp20->getOutput(0), 384, 3, 2, 1, "model.21");
ITensor* inputTensors22[] = { conv21->getOutput(0), conv10->getOutput(0) };
auto cat22 = network->addConcatenation(inputTensors22, 2);
auto bottleneck_csp23 = bottleneckCSP(network, weightMap, *cat22->getOutput(0), 768, 768, 2, false, 1, 0.5, "model.23");
auto bottleneck_csp23 = C3(network, weightMap, *cat22->getOutput(0), 768, 768, 2, false, 1, 0.5, "model.23");
// yolo layer 2
IConvolutionLayer* det2 = network->addConvolutionNd(*bottleneck_csp23->getOutput(0), 3 * (Yolo::CLASS_NUM + 5), DimsHW{ 1, 1 }, weightMap["model.24.m.2.weight"], weightMap["model.24.m.2.bias"]);
@ -218,16 +218,16 @@ ICudaEngine* createEngine_l(unsigned int maxBatchSize, IBuilder* builder, IBuild
/* ------ yolov5 backbone------ */
auto focus0 = focus(network, weightMap, *data, 3, 64, 3, "model.0");
auto conv1 = convBlock(network, weightMap, *focus0->getOutput(0), 128, 3, 2, 1, "model.1");
auto bottleneck_CSP2 = bottleneckCSP(network, weightMap, *conv1->getOutput(0), 128, 128, 3, true, 1, 0.5, "model.2");
auto bottleneck_CSP2 = C3(network, weightMap, *conv1->getOutput(0), 128, 128, 3, true, 1, 0.5, "model.2");
auto conv3 = convBlock(network, weightMap, *bottleneck_CSP2->getOutput(0), 256, 3, 2, 1, "model.3");
auto bottleneck_csp4 = bottleneckCSP(network, weightMap, *conv3->getOutput(0), 256, 256, 9, true, 1, 0.5, "model.4");
auto bottleneck_csp4 = C3(network, weightMap, *conv3->getOutput(0), 256, 256, 9, true, 1, 0.5, "model.4");
auto conv5 = convBlock(network, weightMap, *bottleneck_csp4->getOutput(0), 512, 3, 2, 1, "model.5");
auto bottleneck_csp6 = bottleneckCSP(network, weightMap, *conv5->getOutput(0), 512, 512, 9, true, 1, 0.5, "model.6");
auto bottleneck_csp6 = C3(network, weightMap, *conv5->getOutput(0), 512, 512, 9, true, 1, 0.5, "model.6");
auto conv7 = convBlock(network, weightMap, *bottleneck_csp6->getOutput(0), 1024, 3, 2, 1, "model.7");
auto spp8 = SPP(network, weightMap, *conv7->getOutput(0), 1024, 1024, 5, 9, 13, "model.8");
/* ------ yolov5 head ------ */
auto bottleneck_csp9 = bottleneckCSP(network, weightMap, *spp8->getOutput(0), 1024, 1024, 3, false, 1, 0.5, "model.9");
auto bottleneck_csp9 = C3(network, weightMap, *spp8->getOutput(0), 1024, 1024, 3, false, 1, 0.5, "model.9");
auto conv10 = convBlock(network, weightMap, *bottleneck_csp9->getOutput(0), 512, 1, 1, 1, "model.10");
float *deval = reinterpret_cast<float*>(malloc(sizeof(float) * 512 * 2 * 2));
@ -242,7 +242,7 @@ ICudaEngine* createEngine_l(unsigned int maxBatchSize, IBuilder* builder, IBuild
ITensor* inputTensors12[] = { deconv11->getOutput(0), bottleneck_csp6->getOutput(0) };
auto cat12 = network->addConcatenation(inputTensors12, 2);
auto bottleneck_csp13 = bottleneckCSP(network, weightMap, *cat12->getOutput(0), 1024, 512, 3, false, 1, 0.5, "model.13");
auto bottleneck_csp13 = C3(network, weightMap, *cat12->getOutput(0), 1024, 512, 3, false, 1, 0.5, "model.13");
auto conv14 = convBlock(network, weightMap, *bottleneck_csp13->getOutput(0), 256, 1, 1, 1, "model.14");
Weights deconvwts15{ DataType::kFLOAT, deval, 256 * 2 * 2 };
@ -252,20 +252,20 @@ ICudaEngine* createEngine_l(unsigned int maxBatchSize, IBuilder* builder, IBuild
ITensor* inputTensors16[] = { deconv15->getOutput(0), bottleneck_csp4->getOutput(0) };
auto cat16 = network->addConcatenation(inputTensors16, 2);
auto bottleneck_csp17 = bottleneckCSP(network, weightMap, *cat16->getOutput(0), 512, 256, 3, false, 1, 0.5, "model.17");
auto bottleneck_csp17 = C3(network, weightMap, *cat16->getOutput(0), 512, 256, 3, false, 1, 0.5, "model.17");
// yolo layer 0
IConvolutionLayer* det0 = network->addConvolutionNd(*bottleneck_csp17->getOutput(0), 3 * (Yolo::CLASS_NUM + 5), DimsHW{ 1, 1 }, weightMap["model.24.m.0.weight"], weightMap["model.24.m.0.bias"]);
auto conv18 = convBlock(network, weightMap, *bottleneck_csp17->getOutput(0), 256, 3, 2, 1, "model.18");
ITensor* inputTensors19[] = { conv18->getOutput(0), conv14->getOutput(0) };
auto cat19 = network->addConcatenation(inputTensors19, 2);
auto bottleneck_csp20 = bottleneckCSP(network, weightMap, *cat19->getOutput(0), 512, 512, 3, false, 1, 0.5, "model.20");
auto bottleneck_csp20 = C3(network, weightMap, *cat19->getOutput(0), 512, 512, 3, false, 1, 0.5, "model.20");
//yolo layer 1
IConvolutionLayer* det1 = network->addConvolutionNd(*bottleneck_csp20->getOutput(0), 3 * (Yolo::CLASS_NUM + 5), DimsHW{ 1, 1 }, weightMap["model.24.m.1.weight"], weightMap["model.24.m.1.bias"]);
auto conv21 = convBlock(network, weightMap, *bottleneck_csp20->getOutput(0), 512, 3, 2, 1, "model.21");
ITensor* inputTensors22[] = { conv21->getOutput(0), conv10->getOutput(0) };
auto cat22 = network->addConcatenation(inputTensors22, 2);
auto bottleneck_csp23 = bottleneckCSP(network, weightMap, *cat22->getOutput(0), 1024, 1024, 3, false, 1, 0.5, "model.23");
auto bottleneck_csp23 = C3(network, weightMap, *cat22->getOutput(0), 1024, 1024, 3, false, 1, 0.5, "model.23");
IConvolutionLayer* det2 = network->addConvolutionNd(*bottleneck_csp23->getOutput(0), 3 * (Yolo::CLASS_NUM + 5), DimsHW{ 1, 1 }, weightMap["model.24.m.2.weight"], weightMap["model.24.m.2.bias"]);
auto yolo = addYoLoLayer(network, weightMap, det0, det1, det2);
@ -307,16 +307,16 @@ ICudaEngine* createEngine_x(unsigned int maxBatchSize, IBuilder* builder, IBuild
/* ------ yolov5 backbone------ */
auto focus0 = focus(network, weightMap, *data, 3, 80, 3, "model.0");
auto conv1 = convBlock(network, weightMap, *focus0->getOutput(0), 160, 3, 2, 1, "model.1");
auto bottleneck_CSP2 = bottleneckCSP(network, weightMap, *conv1->getOutput(0), 160, 160, 4, true, 1, 0.5, "model.2");
auto bottleneck_CSP2 = C3(network, weightMap, *conv1->getOutput(0), 160, 160, 4, true, 1, 0.5, "model.2");
auto conv3 = convBlock(network, weightMap, *bottleneck_CSP2->getOutput(0), 320, 3, 2, 1, "model.3");
auto bottleneck_csp4 = bottleneckCSP(network, weightMap, *conv3->getOutput(0), 320, 320, 12, true, 1, 0.5, "model.4");
auto bottleneck_csp4 = C3(network, weightMap, *conv3->getOutput(0), 320, 320, 12, true, 1, 0.5, "model.4");
auto conv5 = convBlock(network, weightMap, *bottleneck_csp4->getOutput(0), 640, 3, 2, 1, "model.5");
auto bottleneck_csp6 = bottleneckCSP(network, weightMap, *conv5->getOutput(0), 640, 640, 12, true, 1, 0.5, "model.6");
auto bottleneck_csp6 = C3(network, weightMap, *conv5->getOutput(0), 640, 640, 12, true, 1, 0.5, "model.6");
auto conv7 = convBlock(network, weightMap, *bottleneck_csp6->getOutput(0), 1280, 3, 2, 1, "model.7");
auto spp8 = SPP(network, weightMap, *conv7->getOutput(0), 1280, 1280, 5, 9, 13, "model.8");
/* ------- yolov5 head ------- */
auto bottleneck_csp9 = bottleneckCSP(network, weightMap, *spp8->getOutput(0), 1280, 1280, 4, false, 1, 0.5, "model.9");
auto bottleneck_csp9 = C3(network, weightMap, *spp8->getOutput(0), 1280, 1280, 4, false, 1, 0.5, "model.9");
auto conv10 = convBlock(network, weightMap, *bottleneck_csp9->getOutput(0), 640, 1, 1, 1, "model.10");
float *deval = reinterpret_cast<float*>(malloc(sizeof(float) * 640 * 2 * 2));
@ -332,7 +332,7 @@ ICudaEngine* createEngine_x(unsigned int maxBatchSize, IBuilder* builder, IBuild
ITensor* inputTensors12[] = { deconv11->getOutput(0), bottleneck_csp6->getOutput(0) };
auto cat12 = network->addConcatenation(inputTensors12, 2);
auto bottleneck_csp13 = bottleneckCSP(network, weightMap, *cat12->getOutput(0), 1280, 640, 4, false, 1, 0.5, "model.13");
auto bottleneck_csp13 = C3(network, weightMap, *cat12->getOutput(0), 1280, 640, 4, false, 1, 0.5, "model.13");
auto conv14 = convBlock(network, weightMap, *bottleneck_csp13->getOutput(0), 320, 1, 1, 1, "model.14");
Weights deconvwts15{ DataType::kFLOAT, deval, 320 * 2 * 2 };
@ -342,20 +342,20 @@ ICudaEngine* createEngine_x(unsigned int maxBatchSize, IBuilder* builder, IBuild
ITensor* inputTensors16[] = { deconv15->getOutput(0), bottleneck_csp4->getOutput(0) };
auto cat16 = network->addConcatenation(inputTensors16, 2);
auto bottleneck_csp17 = bottleneckCSP(network, weightMap, *cat16->getOutput(0), 640, 320, 4, false, 1, 0.5, "model.17");
auto bottleneck_csp17 = C3(network, weightMap, *cat16->getOutput(0), 640, 320, 4, false, 1, 0.5, "model.17");
// yolo layer 0
IConvolutionLayer* det0 = network->addConvolutionNd(*bottleneck_csp17->getOutput(0), 3 * (Yolo::CLASS_NUM + 5), DimsHW{ 1, 1 }, weightMap["model.24.m.0.weight"], weightMap["model.24.m.0.bias"]);
auto conv18 = convBlock(network, weightMap, *bottleneck_csp17->getOutput(0), 320, 3, 2, 1, "model.18");
ITensor* inputTensors19[] = { conv18->getOutput(0), conv14->getOutput(0) };
auto cat19 = network->addConcatenation(inputTensors19, 2);
auto bottleneck_csp20 = bottleneckCSP(network, weightMap, *cat19->getOutput(0), 640, 640, 4, false, 1, 0.5, "model.20");
auto bottleneck_csp20 = C3(network, weightMap, *cat19->getOutput(0), 640, 640, 4, false, 1, 0.5, "model.20");
// yolo layer 1
IConvolutionLayer* det1 = network->addConvolutionNd(*bottleneck_csp20->getOutput(0), 3 * (Yolo::CLASS_NUM + 5), DimsHW{ 1, 1 }, weightMap["model.24.m.1.weight"], weightMap["model.24.m.1.bias"]);
auto conv21 = convBlock(network, weightMap, *bottleneck_csp20->getOutput(0), 640, 3, 2, 1, "model.21");
ITensor* inputTensors22[] = { conv21->getOutput(0), conv10->getOutput(0) };
auto cat22 = network->addConcatenation(inputTensors22, 2);
auto bottleneck_csp23 = bottleneckCSP(network, weightMap, *cat22->getOutput(0), 1280, 1280, 4, false, 1, 0.5, "model.23");
auto bottleneck_csp23 = C3(network, weightMap, *cat22->getOutput(0), 1280, 1280, 4, false, 1, 0.5, "model.23");
// yolo layer 2
IConvolutionLayer* det2 = network->addConvolutionNd(*bottleneck_csp23->getOutput(0), 3 * (Yolo::CLASS_NUM + 5), DimsHW{ 1, 1 }, weightMap["model.24.m.2.weight"], weightMap["model.24.m.2.bias"]);