A minimal PyTorch implementation of YOLOv4.
-
Paper Yolo v4: https://arxiv.org/abs/2004.10934
-
Source code:https://github.com/AlexeyAB/darknet
-
More details: http://pjreddie.com/darknet/yolo/
-
Inference
-
Train
- Mosaic
├── README.md
├── dataset.py dataset
├── demo.py demo to run pytorch --> tool/darknet2pytorch
├── demo_darknet2onnx.py tool to convert into onnx --> tool/darknet2pytorch
├── demo_pytorch2onnx.py tool to convert into onnx
├── models.py model for pytorch
├── train.py train models.py
├── cfg.py cfg.py for train
├── cfg cfg --> darknet2pytorch
├── data
├── weight --> darknet2pytorch
├── tool
│ ├── camera.py a demo camera
│ ├── coco_annotation.py coco dataset generator
│ ├── config.py
│ ├── darknet2pytorch.py
│ ├── region_loss.py
│ ├── utils.py
│ └── yolo_layer.py
- baidu(https://pan.baidu.com/s/1dAGEW8cm-dqK14TbhhVetA Extraction code:dm5b)
- google(https://drive.google.com/open?id=1cewMfusmPjYWbrnuJRuKhPMwRe_b9PaT)
you can use darknet2pytorch to convert it yourself, or download my converted model.
- baidu
- yolov4.pth(https://pan.baidu.com/s/1ZroDvoGScDgtE1ja_QqJVw Extraction code:xrq9)
- yolov4.conv.137.pth(https://pan.baidu.com/s/1ovBie4YyVQQoUrC3AY0joA Extraction code:kcel)
- google
- yolov4.pth(https://drive.google.com/open?id=1wv_LiFeCRYwtpkqREPeI13-gPELBDwuJ)
- yolov4.conv.137.pth(https://drive.google.com/open?id=1fcbR0bWzYfIEdLJPzOsn4R5mlvR6IQyA)
use yolov4 to train your own data
-
Download weight
-
Transform data
For coco dataset,you can use tool/coco_annotation.py.
# train.txt image_path1 x1,y1,x2,y2,id x1,y1,x2,y2,id x1,y1,x2,y2,id ... image_path2 x1,y1,x2,y2,id x1,y1,x2,y2,id x1,y1,x2,y2,id ... ... ... -
Train
you can set parameters in cfg.py.
python train.py -g [GPU_ID] -dir [Dataset direction] ...
2.1 Performance on MS COCO dataset (using pretrained DarknetWeights from https://github.com/AlexeyAB/darknet)
ONNX and TensorRT models are converted from Pytorch (TianXiaomo): Pytorch->ONNX->TensorRT. See following sections for more details of conversions.
- val2017 dataset (input size: 416x416)
| Model type | AP | AP50 | AP75 | APS | APM | APL |
|---|---|---|---|---|---|---|
| DarkNet (YOLOv4 paper) | 0.471 | 0.710 | 0.510 | 0.278 | 0.525 | 0.636 |
| Pytorch (TianXiaomo) | 0.466 | 0.704 | 0.505 | 0.267 | 0.524 | 0.629 |
| TensorRT FP32 + BatchedNMSPlugin | 0.472 | 0.708 | 0.511 | 0.273 | 0.530 | 0.637 |
| TensorRT FP16 + BatchedNMSPlugin | 0.472 | 0.708 | 0.511 | 0.273 | 0.530 | 0.636 |
- testdev2017 dataset (input size: 416x416)
| Model type | AP | AP50 | AP75 | APS | APM | APL |
|---|---|---|---|---|---|---|
| DarkNet (YOLOv4 paper) | 0.412 | 0.628 | 0.443 | 0.204 | 0.444 | 0.560 |
| Pytorch (TianXiaomo) | 0.404 | 0.615 | 0.436 | 0.196 | 0.438 | 0.552 |
| TensorRT FP32 + BatchedNMSPlugin | 0.412 | 0.625 | 0.445 | 0.200 | 0.446 | 0.564 |
| TensorRT FP16 + BatchedNMSPlugin | 0.412 | 0.625 | 0.445 | 0.200 | 0.446 | 0.563 |
Image input size is NOT restricted in 320 * 320, 416 * 416, 512 * 512 and 608 * 608.
You can adjust your input sizes for a different input ratio, for example: 320 * 608.
Larger input size could help detect smaller targets, but may be slower and GPU memory exhausting.
height = 320 + 96 * n, n in {0, 1, 2, 3, ...}
width = 320 + 96 * m, m in {0, 1, 2, 3, ...}-
Load the pretrained darknet model and darknet weights to do the inference (image size is configured in cfg file already)
python demo.py -cfgfile <cfgFile> -weightfile <weightFile> -imgfile <imgFile>
-
Load pytorch weights (pth file) to do the inference
python models.py <num_classes> <weightfile> <imgfile> <IN_IMAGE_H> <IN_IMAGE_W> <namefile(optional)>
-
Load converted ONNX file to do inference (See section 3 and 4)
-
Load converted TensorRT engine file to do inference (See section 5)
There are 2 inference outputs.
- One is locations of bounding boxes, its shape is
[batch, num_boxes, 1, 4]which represents x1, y1, x2, y2 of each bounding box. - The other one is scores of bounding boxes which is of shape
[batch, num_boxes, num_classes]indicating scores of all classes for each bounding box.
Until now, still a small piece of post-processing including NMS is required. We are trying to minimize time and complexity of post-processing.
-
This script is to convert the official pretrained darknet model into ONNX
-
Pytorch version Recommended:
- Pytorch 1.4.0 for TensorRT 7.0 and higher
- Pytorch 1.5.0 and 1.6.0 for TensorRT 7.1.2 and higher
-
Install onnxruntime
pip install onnxruntime
-
Run python script to generate ONNX model and run the demo
python demo_darknet2onnx.py <cfgFile> <namesFile> <weightFile> <imageFile> <batchSize>
- Positive batch size will generate ONNX model of static batch size, otherwise, batch size will be dynamic
- Dynamic batch size will generate only one ONNX model
- Static batch size will generate 2 ONNX models, one is for running the demo (batch_size=1)
-
You can convert your trained pytorch model into ONNX using this script
-
Pytorch version Recommended:
- Pytorch 1.4.0 for TensorRT 7.0 and higher
- Pytorch 1.5.0 and 1.6.0 for TensorRT 7.1.2 and higher
-
Install onnxruntime
pip install onnxruntime
-
Run python script to generate ONNX model and run the demo
python demo_pytorch2onnx.py <weight_file> <image_path> <batch_size> <n_classes> <IN_IMAGE_H> <IN_IMAGE_W>
For example:
python demo_pytorch2onnx.py yolov4.pth dog.jpg 8 80 416 416
- Positive batch size will generate ONNX model of static batch size, otherwise, batch size will be dynamic
- Dynamic batch size will generate only one ONNX model
- Static batch size will generate 2 ONNX models, one is for running the demo (batch_size=1)
- TensorRT version Recommended: 7.0, 7.1
-
Run the following command to convert YOLOv4 ONNX model into TensorRT engine
trtexec --onnx=<onnx_file> --explicitBatch --saveEngine=<tensorRT_engine_file> --workspace=<size_in_megabytes> --fp16
- Note: If you want to use int8 mode in conversion, extra int8 calibration is needed.
-
Run the following command to convert YOLOv4 ONNX model into TensorRT engine
trtexec --onnx=<onnx_file> \ --minShapes=input:<shape_of_min_batch> --optShapes=input:<shape_of_opt_batch> --maxShapes=input:<shape_of_max_batch> \ --workspace=<size_in_megabytes> --saveEngine=<engine_file> --fp16
-
For example:
trtexec --onnx=yolov4_-1_3_320_512_dynamic.onnx \ --minShapes=input:1x3x320x512 --optShapes=input:4x3x320x512 --maxShapes=input:8x3x320x512 \ --workspace=2048 --saveEngine=yolov4_-1_3_320_512_dynamic.engine --fp16
python demo_trt.py <tensorRT_engine_file> <input_image> <input_H> <input_W>-
This demo here only works when batchSize is dynamic (1 should be within dynamic range) or batchSize=1, but you can update this demo a little for other dynamic or static batch sizes.
-
Note1: input_H and input_W should agree with the input size in the original ONNX file.
-
Note2: extra NMS operations are needed for the tensorRT output. This demo uses python NMS code from
tool/utils.py.
-
First:Conversion to ONNX
tensorflow >=2.0
1: Thanks:github:https://github.com/onnx/onnx-tensorflow
2: Run git clone https://github.com/onnx/onnx-tensorflow.git && cd onnx-tensorflow Run pip install -e .
Note:Errors will occur when using "pip install onnx-tf", at least for me,it is recommended to use source code installation
- Compile the DeepStream Nvinfer Plugin
cd DeepStream
make
- Build a TRT Engine.
For single batch,
trtexec --onnx=<onnx_file> --explicitBatch --saveEngine=<tensorRT_engine_file> --workspace=<size_in_megabytes> --fp16
For multi-batch,
trtexec --onnx=<onnx_file> --explicitBatch --shapes=input:Xx3xHxW --optShapes=input:Xx3xHxW --maxShapes=input:Xx3xHxW --minShape=input:1x3xHxW --saveEngine=<tensorRT_engine_file> --fp16
Note :The maxShapes could not be larger than model original shape.
- Write the deepstream config file for the TRT Engine.
Reference:
- https://github.com/eriklindernoren/PyTorch-YOLOv3
- https://github.com/marvis/pytorch-caffe-darknet-convert
- https://github.com/marvis/pytorch-yolo3
@article{yolov4,
title={YOLOv4: YOLOv4: Optimal Speed and Accuracy of Object Detection},
author={Alexey Bochkovskiy, Chien-Yao Wang, Hong-Yuan Mark Liao},
journal = {arXiv},
year={2020}
}