README.md 15.8 KB
Newer Older
1 2 3 4
English | [简体中文](README_cn.md)

# PP-YOLOE

5 6 7 8 9 10 11 12 13
## Latest News
- Release PP-YOLOE+ model: **(2022.08)**
  - Pre training model using large-scale data set obj365
  - In the backbone, add the alpha parameter to the block branch
  - Optimize the end-to-end inference speed and improve the training convergence speed

## Legacy model
- Please refer to:[PP-YOLOE 2022.03](./README_legacy.md) for details

14 15 16 17 18 19 20
## Table of Contents
- [Introduction](#Introduction)
- [Model Zoo](#Model-Zoo)
- [Getting Start](#Getting-Start)
- [Appendix](#Appendix)

## Introduction
21
PP-YOLOE is an excellent single-stage anchor-free model based on PP-YOLOv2, surpassing a variety of popular YOLO models. PP-YOLOE has a series of models, named s/m/l/x, which are configured through width multiplier and depth multiplier. PP-YOLOE avoids using special operators, such as Deformable Convolution or Matrix NMS, to be deployed friendly on various hardware. For more details, please refer to our [report](https://arxiv.org/abs/2203.16250).
22 23

<div align="center">
24
  <img src="../../docs/images/ppyoloe_plus_map_fps.png" width=500 />
25 26
</div>

27
PP-YOLOE+_l achieves 53.3 mAP on COCO test-dev2017 dataset with 78.1 FPS on Tesla V100. While using TensorRT FP16, PP-YOLOE+_l can be further accelerated to 149.2 FPS. PP-YOLOE+_s/m/x also have excellent accuracy and speed performance, which can be found in [Model Zoo](#Model-Zoo)
28 29 30 31 32

PP-YOLOE is composed of following methods:
- Scalable backbone and neck
- [Task Alignment Learning](https://arxiv.org/abs/2108.07755)
- Efficient Task-aligned head with [DFL](https://arxiv.org/abs/2006.04388) and [VFL](https://arxiv.org/abs/2008.13367)
33
- [SiLU(Swish) activation function](https://arxiv.org/abs/1710.05941)
34 35

## Model Zoo
36 37 38 39 40 41
|     Model      | Epoch | GPU number | images/GPU |  backbone  | input shape | Box AP<sup>val<br>0.5:0.95 | Box AP<sup>test<br>0.5:0.95 | Params(M) | FLOPs(G) | V100 FP32(FPS) | V100 TensorRT FP16(FPS) |                                       download                                       |                   config                    |
|:--------------:|:-----:|:-------:|:----------:|:----------:| :-------:|:--------------------------:|:---------------------------:|:---------:|:--------:|:---------------:| :---------------------: |:------------------------------------------------------------------------------------:|:-------------------------------------------:|
|  PP-YOLOE+_s   |  80   |     8      |     8      | cspresnet-s |     640     |            43.7            |            43.9             |   7.93    |  17.36   |   208.3   |  333.3   | [model](https://paddledet.bj.bcebos.com/models/ppyoloe_plus_crn_s_80e_coco.pdparams) | [config](./ppyoloe_plus_crn_s_80e_coco.yml) |
|  PP-YOLOE+_m   |  80   |     8      |     8      | cspresnet-m |     640     |            49.8            |            50.0             |   23.43   |  49.91   |   123.4   |  208.3   | [model](https://paddledet.bj.bcebos.com/models/ppyoloe_plus_crn_m_80e_coco.pdparams) | [config](./ppyoloe_plus_crn_m_80e_coco.yml) |
|  PP-YOLOE+_l   |  80   |     8      |     8      | cspresnet-l |     640     |            52.9            |            53.3             |   52.20   |  110.07  |   78.1    |  149.2   | [model](https://paddledet.bj.bcebos.com/models/ppyoloe_plus_crn_l_80e_coco.pdparams) | [config](./ppyoloe_plus_crn_l_80e_coco.yml) |
|  PP-YOLOE+_x   |  80   |     8      |     8      | cspresnet-x |     640     |            54.7            |            54.9             |   98.42   |  206.59  |   45.0    |   95.2   | [model](https://paddledet.bj.bcebos.com/models/ppyoloe_plus_crn_x_80e_coco.pdparams) | [config](./ppyoloe_plus_crn_x_80e_coco.yml) |
42

43 44

### Comprehensive Metrics
45 46 47 48 49 50
|          Model           | Epoch | AP<sup>0.5:0.95 | AP<sup>0.5 | AP<sup>0.75  | AP<sup>small | AP<sup>medium | AP<sup>large | AR<sup>small | AR<sup>medium | AR<sup>large |
|:------------------------:|:-----:|:---------------:|:----------:|:------------:|:------------:| :-----------: |:------------:|:------------:|:-------------:|:------------:|
|       PP-YOLOE+_s        |  80   |      43.7       |    60.6    |     47.9     |     26.5     |     47.5      |     59.0     |     46.7     |     71.4      |     81.7     |
|       PP-YOLOE+_m        |  80   |      49.8       |    67.1    |     54.5     |     31.8     |     53.9      |     66.2     |     53.3     |     75.0      |     84.6     |
|       PP-YOLOE+_l        |  80   |      52.9       |    70.1    |     57.9     |     35.2     |     57.5      |     69.1     |     56.0     |     77.9      |     86.9     |
|       PP-YOLOE+_x        |  80   |      54.7       |    72.0    |     59.9     |     37.9     |     59.3      |     70.4     |     57.0     |     78.7      |     87.2     |
51

52

53 54 55 56 57 58 59 60
### End-to-end Speed
|    Model    | AP<sup>0.5:0.95 | TRT-FP32(fps) | TRT-FP16(fps) |
|:-----------:|:---------------:|:-------------:|:-------------:|
| PP-YOLOE+_s |      43.7       |     44.44     |     47.85     |
| PP-YOLOE+_m |      49.8       |     39.06     |     43.86     |
| PP-YOLOE+_l |      52.9       |     34.01     |     42.02     |
| PP-YOLOE+_x |      54.7       |     26.88     |     36.76     |

61 62
**Notes:**

63
- PP-YOLOE is trained on COCO train2017 dataset and evaluated on val2017 & test-dev2017 dataset.
64
- The model weights in the table of Comprehensive Metrics are **the same as** that in the original Model Zoo, and evaluated on **val2017**.
65
- PP-YOLOE used 8 GPUs for mixed precision training, if **GPU number** or **mini-batch size** is changed, **learning rate** should be adjusted according to the formula **lr<sub>new</sub> = lr<sub>default</sub> * (batch_size<sub>new</sub> * GPU_number<sub>new</sub>) / (batch_size<sub>default</sub> * GPU_number<sub>default</sub>)**.
66 67
- PP-YOLOE inference speed is tesed on single Tesla V100 with batch size as 1, **CUDA 10.2**, **CUDNN 7.6.5**, **TensorRT 6.0.1.8** in TensorRT mode.
- Refer to [Speed testing](#Speed-testing) to reproduce the speed testing results of PP-YOLOE.
68
- If you set `--run_benchmark=True`,you should install these dependencies at first, `pip install pynvml psutil GPUtil`.
69
- End-to-end speed test includes pre-processing + inference + post-processing and NMS time, using **Intel(R) Xeon(R) Gold 5117 CPU @ 2.00GHz**, **single Tesla V100**, **CUDA 11.2**, **CUDNN 8.2.0**, **TensorRT 8.0.1.6**.
70 71 72 73 74 75 76 77 78 79 80 81

### Feature Models

The PaddleDetection team provides configs and weights of various feature detection models based on PP-YOLOE, which users can download for use:

|Scenarios | Related Datasets | Links|
| :--------: | :---------: | :------: |
|Pedestrian Detection | CrowdHuman | [pphuman](../pphuman) |
|Vehicle Detection | BDD100K, UA-DETRAC | [ppvehicle](../ppvehicle) |
|Small Object Detection | VisDrone | [visdrone](../visdrone) |


82 83
## Getting Start

84
### Training
85

86
Training PP-YOLOE+ on 8 GPUs with following command
87 88

```bash
89
python -m paddle.distributed.launch --gpus 0,1,2,3,4,5,6,7 tools/train.py -c configs/ppyoloe/ppyoloe_plus_crn_l_80e_coco.yml --eval --amp
90 91
```

W
Wenyu 已提交
92
**Notes:**
93
- If you need to evaluate while training, please add `--eval`.
94
- PP-YOLOE+ supports mixed precision training, please add `--amp`.
W
Wenyu 已提交
95 96
- PaddleDetection supports multi-machine distribued training, you can refer to [DistributedTraining tutorial](../../docs/DistributedTraining_en.md).

97

98
### Evaluation
99

100
Evaluating PP-YOLOE+ on COCO val2017 dataset in single GPU with following commands:
101 102

```bash
103
CUDA_VISIBLE_DEVICES=0 python tools/eval.py -c configs/ppyoloe/ppyoloe_plus_crn_l_80e_coco.yml -o weights=https://paddledet.bj.bcebos.com/models/ppyoloe_plus_crn_l_80e_coco.pdparams
104 105 106 107
```

For evaluation on COCO test-dev2017 dataset, please download COCO test-dev2017 dataset from [COCO dataset download](https://cocodataset.org/#download) and decompress to COCO dataset directory and configure `EvalDataset` like `configs/ppyolo/ppyolo_test.yml`.

108
### Inference
109 110 111 112 113

Inference images in single GPU with following commands, use `--infer_img` to inference a single image and `--infer_dir` to inference all images in the directory.

```bash
# inference single image
114
CUDA_VISIBLE_DEVICES=0 python tools/infer.py -c configs/ppyoloe/ppyoloe_plus_crn_l_80e_coco.yml -o weights=https://paddledet.bj.bcebos.com/models/ppyoloe_plus_crn_l_80e_coco.pdparams --infer_img=demo/000000014439_640x640.jpg
115 116

# inference all images in the directory
117
CUDA_VISIBLE_DEVICES=0 python tools/infer.py -c configs/ppyoloe/ppyoloe_plus_crn_l_80e_coco.yml -o weights=https://paddledet.bj.bcebos.com/models/ppyoloe_plus_crn_l_80e_coco.pdparams --infer_dir=demo
118 119
```

120
### Exporting models
121

122
For deployment on GPU or speed testing, model should be first exported to inference model using `tools/export_model.py`.
W
Wenyu 已提交
123

124
**Exporting PP-YOLOE+ for Paddle Inference without TensorRT**, use following command
125 126

```bash
127
python tools/export_model.py -c configs/ppyoloe/ppyoloe_plus_crn_l_80e_coco.yml -o weights=https://paddledet.bj.bcebos.com/models/ppyoloe_plus_crn_l_80e_coco.pdparams
128 129
```

130
**Exporting PP-YOLOE+ for Paddle Inference with TensorRT** for better performance, use following command with extra `-o trt=True` setting.
131 132

```bash
133
python tools/export_model.py -c configs/ppyoloe/ppyoloe_plus_crn_l_80e_coco.yml -o weights=https://paddledet.bj.bcebos.com/models/ppyoloe_plus_crn_l_80e_coco.pdparams trt=True
134 135
```

136
If you want to export PP-YOLOE model to **ONNX format**, use following command refer to [PaddleDetection Model Export as ONNX Format Tutorial](../../deploy/EXPORT_ONNX_MODEL_en.md).
137 138

```bash
139
# export inference model
140
python tools/export_model.py -c configs/ppyoloe/ppyoloe_plus_crn_l_80e_coco.yml --output_dir=output_inference -o weights=https://paddledet.bj.bcebos.com/models/ppyoloe_plus_crn_l_80e_coco.pdparams trt=True
141

142 143 144 145
# install paddle2onnx
pip install paddle2onnx

# convert to onnx
146
paddle2onnx --model_dir output_inference/ppyoloe_plus_crn_l_80e_coco --model_filename model.pdmodel --params_filename model.pdiparams --opset_version 11 --save_file ppyoloe_plus_crn_l_80e_coco.onnx
147 148 149

```

150 151 152 153 154 155 156
**Notes:** ONNX model only supports batch_size=1 now

### Speed testing

For fair comparison, the speed in [Model Zoo](#Model-Zoo) do not contains the time cost of data reading and post-processing(NMS), which is same as [YOLOv4(AlexyAB)](https://github.com/AlexeyAB/darknet) in testing method. Thus, you should export model with extra `-o exclude_nms=True` setting.

**Using Paddle Inference without TensorRT** to test speed, run following command
157 158 159

```bash
# export inference model
160
python tools/export_model.py -c configs/ppyoloe/ppyoloe_plus_crn_l_80e_coco.yml -o weights=https://paddledet.bj.bcebos.com/models/ppyoloe_plus_crn_l_80e_coco.pdparams exclude_nms=True
161

162
# speed testing with run_benchmark=True
163
CUDA_VISIBLE_DEVICES=0 python deploy/python/infer.py --model_dir=output_inference/ppyoloe_plus_crn_l_80e_coco --image_file=demo/000000014439_640x640.jpg --run_mode=paddle --device=gpu --run_benchmark=True
164
```
165

166 167 168 169
**Using Paddle Inference with TensorRT** to test speed, run following command

```bash
# export inference model with trt=True
170
python tools/export_model.py -c configs/ppyoloe/ppyoloe_plus_crn_l_80e_coco.yml -o weights=https://paddledet.bj.bcebos.com/models/ppyoloe_plus_crn_l_80e_coco.pdparams exclude_nms=True trt=True
171 172

# speed testing with run_benchmark=True,run_mode=trt_fp32/trt_fp16
173
CUDA_VISIBLE_DEVICES=0 python deploy/python/infer.py --model_dir=output_inference/ppyoloe_plus_crn_l_80e_coco --image_file=demo/000000014439_640x640.jpg --run_mode=trt_fp16 --device=gpu --run_benchmark=True
174 175 176

```

177 178 179 180
**Using TensorRT Inference with ONNX** to test speed, run following command

```bash
# export inference model with trt=True
181
python tools/export_model.py -c configs/ppyoloe/ppyoloe_plus_crn_s_80e_coco.yml -o weights=https://paddledet.bj.bcebos.com/models/ppyoloe_plus_crn_s_80e_coco.pdparams exclude_nms=True trt=True
182 183

# convert to onnx
184
paddle2onnx --model_dir output_inference/ppyoloe_plus_crn_s_80e_coco --model_filename model.pdmodel --params_filename model.pdiparams --opset_version 12 --save_file ppyoloe_plus_crn_s_80e_coco.onnx
185 186

# trt inference using fp16 and batch_size=1
187
trtexec --onnx=./ppyoloe_plus_crn_s_80e_coco.onnx --saveEngine=./ppyoloe_s_bs1.engine --workspace=1024 --avgRuns=1000 --shapes=image:1x3x640x640,scale_factor:1x2 --fp16
188 189

# trt inference using fp16 and batch_size=32
190
trtexec --onnx=./ppyoloe_plus_crn_s_80e_coco.onnx --saveEngine=./ppyoloe_s_bs32.engine --workspace=1024 --avgRuns=1000 --shapes=image:32x3x640x640,scale_factor:32x2 --fp16
191 192 193 194 195 196 197 198 199

# Using the above script, T4 and tensorrt 7.2 machine, the speed of PPYOLOE-s model is as follows,

# batch_size=1, 2.80ms, 357fps
# batch_size=32, 67.69ms, 472fps

```


200 201 202 203 204 205
### Deployment

PP-YOLOE can be deployed by following approches:
  - Paddle Inference [Python](../../deploy/python) & [C++](../../deploy/cpp)
  - [Paddle-TensorRT](../../deploy/TENSOR_RT.md)
  - [PaddleServing](https://github.com/PaddlePaddle/Serving)
W
Wenyu 已提交
206
  - [PaddleSlim](../slim)
207 208 209 210 211 212 213 214

Next, we will introduce how to use Paddle Inference to deploy PP-YOLOE models in TensorRT FP16 mode.

First, refer to [Paddle Inference Docs](https://www.paddlepaddle.org.cn/inference/master/user_guides/download_lib.html#python), download and install packages corresponding to CUDA, CUDNN and TensorRT version.

Then, Exporting PP-YOLOE for Paddle Inference **with TensorRT**, use following command.

```bash
215
python tools/export_model.py -c configs/ppyoloe/ppyoloe_plus_crn_l_80e_coco.yml -o weights=https://paddledet.bj.bcebos.com/models/ppyoloe_plus_crn_l_80e_coco.pdparams trt=True
216 217 218 219 220 221
```

Finally, inference in TensorRT FP16 mode.

```bash
# inference single image
222
CUDA_VISIBLE_DEVICES=0 python deploy/python/infer.py --model_dir=output_inference/ppyoloe_plus_crn_l_80e_coco --image_file=demo/000000014439_640x640.jpg --device=gpu --run_mode=trt_fp16
223 224

# inference all images in the directory
225
CUDA_VISIBLE_DEVICES=0 python deploy/python/infer.py --model_dir=output_inference/ppyoloe_plus_crn_l_80e_coco --image_dir=demo/ --device=gpu  --run_mode=trt_fp16
226 227 228

```

229
**Notes:**
230 231 232 233
- TensorRT will perform optimization for the current hardware platform according to the definition of the network, generate an inference engine and serialize it into a file. This inference engine is only applicable to the current hardware hardware platform. If your hardware and software platform has not changed, you can set `use_static=True` in [enable_tensorrt_engine](https://github.com/PaddlePaddle/PaddleDetection/blob/release/2.4/deploy/python/infer.py#L660). In this way, the serialized file generated will be saved in the `output_inference` folder, and the saved serialized file will be loaded the next time when TensorRT is executed.
- PaddleDetection release/2.4 and later versions will support NMS calling TensorRT, which requires PaddlePaddle release/2.3 and later versions.

### Other Datasets
W
Wenyu 已提交
234 235 236 237 238 239 240

Model | AP | AP<sub>50</sub>
---|---|---
[YOLOX](https://github.com/Megvii-BaseDetection/YOLOX) | 22.6 | 37.5
[YOLOv5](https://github.com/ultralytics/yolov5) | 26.0 | 42.7
**PP-YOLOE** | **30.5** | **46.4**

241
**Notes**
W
Wenyu 已提交
242 243 244 245 246
- Here, we use [VisDrone](https://github.com/VisDrone/VisDrone-Dataset) dataset, and to detect 9 objects including `person, bicycles, car, van, truck, tricyle, awning-tricyle, bus, motor`.
- Above models trained using official default config, and load pretrained parameters on COCO dataset.
- *Due to the limited time, more verification results will be supplemented in the future. You are also welcome to contribute to PP-YOLOE*


247 248 249 250 251 252 253 254 255 256 257
## Appendix

Ablation experiments of PP-YOLOE.

| NO.  |        Model                 | Box AP<sup>val</sup> | Params(M) | FLOPs(G) | V100 FP32 FPS |
| :--: | :---------------------------: | :------------------: | :-------: | :------: | :-----------: |
|  A   | PP-YOLOv2          |         49.1         |   54.58   |  115.77   |     68.9     |
|  B   | A + Anchor-free    |         48.8         |   54.27   |  114.78   |     69.8     |
|  C   | B + CSPRepResNet   |         49.5         |   47.42   |  101.87   |     85.5     |
|  D   | C + TAL            |         50.4         |   48.32   |  104.75   |     84.0     |
|  E   | D + ET-Head        |         50.9         |   52.20   |  110.07   |     78.1     |