-`CBResNet` stands for `Cascade-Faster-RCNN-CBResNet200vd-FPN`, which has highest mAP on COCO as 53.3% in PaddleDetection models
-`Cascade-Faster-RCNN` stands for `Cascade-Faster-RCNN-ResNet50vd-DCN`, which has been optimized to 20 FPS inference speed when COCO mAP as 47.8%
-The enhanced `YOLOv3-ResNet50vd-DCN` is 10.6 absolute percentage points higher than paper on COCO mAP, and inference speed is nearly 70% faster than the darknet framework
-All these models can be get in [Model Zoo](#Model-Zoo)
The following is the relationship between COCO mAP and FPS on Tesla V100 of SOTA object detecters and PP-YOLO, which is faster and has better performance than YOLOv4, and reached mAP(0.5:0.95) as 45.2% on COCO test2019 dataset and 72.9 FPS on single Test V100. Please refer to [PP-YOLO](configs/ppyolo/README.md) for details.
-[Pretrained models for pedestrian detection](docs/featured_model/CONTRIB.md)
-[Pretrained models for vehicle detection](docs/featured_model/CONTRIB.md)
-[YOLOv3 enhanced model](docs/featured_model/YOLOv3_ENHANCEMENT.md): Compared to MAP of 33.0% in paper, enhanced YOLOv3 reaches the MAP of 43.6%, and inference speed is improved as well
-[PP-YOLO](configs/ppyolo/README.md): PP-YOLO reeached mAP as 45.3% on COCO dataset，and 72.9 FPS on single Tesla V100
-[Best single model of Open Images 2019-Object Detction](docs/featured_model/champion_model/OIDV5_BASELINE_MODEL.md)
-[Practical Server-side detection method](configs/rcnn_enhance/README_en.md): Inference speed on single V100 GPU can reach 20FPS when COCO mAP is 47.8%.
-[Large-scale practical object detection models](docs/featured_model/LARGE_SCALE_DET_MODEL_en.md): Large-scale practical server-side detection pretrained models with 676 categories are provided for most application scenarios, which can be used not only for direct inference but also finetuning on other datasets.
## License
PaddleDetection is released under the [Apache 2.0 license](LICENSE).
## Updates
v0.4.0 was released at `05/2020`, add PP-YOLO, TTFNet, HTC, ACFPN, etc. And add BlaceFace face landmark detection model, add a series of optimized SSDLite models on mobile side, add data augmentations GridMask and RandomErasing, add Matrix NMS and EMA training, and improved ease of use, fix many known bugs, etc.
Please refer to [版本更新文档](docs/CHANGELOG.md) for details.
## Contributing
Contributions are highly welcomed and we would really appreciate your feedback!!
<aname="vd">[1]</a> [ResNet-vd](https://arxiv.org/pdf/1812.01187) models offer much improved accuracy with negligible performance cost.
**NOTE:** ✓ for config file and pretrain model provided in [Model Zoo](docs/MODEL_ZOO.md), ✗ for not provided but is supported generally.
More models:
- EfficientDet
- FCOS
- CornerNet-Squeeze
- YOLOv4
- PP-YOLO
More Backbones:
- DarkNet
- VGG
- GCNet
- CBNet
Advanced Features:
- [x] **Synchronized Batch Norm**
- [x] **Group Norm**
- [x] **Modulated Deformable Convolution**
- [x] **Deformable PSRoI Pooling**
- [x] **Non-local and GCNet**
**NOTE:** Synchronized batch normalization can only be used on multiple GPU devices, can not be used on CPU devices or single GPU device.
The following is the relationship between COCO mAP and FPS on Tesla V100 of representative models of each architectures and backbones.
<divalign="center">
<imgsrc="docs/images/map_fps.png"width=800/>
</div>
**NOTE:**
-`CBResNet` stands for `Cascade-Faster-RCNN-CBResNet200vd-FPN`, which has highest mAP on COCO as 53.3% in PaddleDetection models
-`Cascade-Faster-RCNN` stands for `Cascade-Faster-RCNN-ResNet50vd-DCN`, which has been optimized to 20 FPS inference speed when COCO mAP as 47.8%
- The enhanced `YOLOv3-ResNet50vd-DCN` is 10.6 absolute percentage points higher than paper on COCO mAP, and inference speed is nearly 70% faster than the darknet framework
- All these models can be get in [Model Zoo](#Model-Zoo)
The following is the relationship between COCO mAP and FPS on Tesla V100 of SOTA object detecters and PP-YOLO, which is faster and has better performance than YOLOv4, and reached mAP(0.5:0.95) as 45.2% on COCO test2019 dataset and 72.9 FPS on single Test V100. Please refer to [PP-YOLO](configs/ppyolo/README.md) for details.
-[Pretrained models for pedestrian detection](docs/featured_model/CONTRIB.md)
-[Pretrained models for vehicle detection](docs/featured_model/CONTRIB.md)
-[YOLOv3 enhanced model](docs/featured_model/YOLOv3_ENHANCEMENT.md): Compared to MAP of 33.0% in paper, enhanced YOLOv3 reaches the MAP of 43.6%, and inference speed is improved as well
-[PP-YOLO](configs/ppyolo/README.md): PP-YOLO reeached mAP as 45.3% on COCO dataset，and 72.9 FPS on single Tesla V100
-[Best single model of Open Images 2019-Object Detction](docs/featured_model/champion_model/OIDV5_BASELINE_MODEL.md)
-[Practical Server-side detection method](configs/rcnn_enhance/README_en.md): Inference speed on single V100 GPU can reach 20FPS when COCO mAP is 47.8%.
-[Large-scale practical object detection models](docs/featured_model/LARGE_SCALE_DET_MODEL_en.md): Large-scale practical server-side detection pretrained models with 676 categories are provided for most application scenarios, which can be used not only for direct inference but also finetuning on other datasets.
## License
PaddleDetection is released under the [Apache 2.0 license](LICENSE).
## Updates
v0.4.0 was released at `05/2020`, add PP-YOLO, TTFNet, HTC, ACFPN, etc. And add BlaceFace face landmark detection model, add a series of optimized SSDLite models on mobile side, add data augmentations GridMask and RandomErasing, add Matrix NMS and EMA training, and improved ease of use, fix many known bugs, etc.
Please refer to [版本更新文档](docs/CHANGELOG.md) for details.
## Contributing
Contributions are highly welcomed and we would really appreciate your feedback!!
[PP-YOLO](https://arxiv.org/abs/2007.12099) is a optimized model based on YOLOv3 in PaddleDetection，whose performance(mAP on COCO) and inference spped are better than [YOLOv4](https://arxiv.org/abs/2004.10934)，PaddlePaddle 1.8.4(will release in mid-August 202) or [Daily Version](https://www.paddlepaddle.org.cn/documentation/docs/zh/install/Tables.html#whl-dev) is required to run this PP-YOLO。
PP-YOLO reached mmAP(IoU=0.5:0.95) as 45.2% on COCO test-dev2017 dataset, and inference speed of FP32 on single V100 is 72.9 FPS, inference speed of FP16 with TensorRT on single V100 is 155.6 FPS.
- PP-YOLO is trained on COCO train2017 datast and evaluated on test-dev2017 dataset，`Box AP` is evaluation results as `mAP(IoU=0.5:0.95)`.
- PP-YOLO used 8 GPUs for training and mini-batch size as 24 on each GPU, if GPU number and mini-batch size is changed, learning rate and iteration times should be adjusted according [FAQ](../../docs/FAQ.md).
- PP-YOLO inference speed is tesed on single Tesla V100 with batch size as 1, CUDA 10.2, CUDNN 7.5.1, TensorRT 5.1.2.2 in TensorRT mode.
- PP-YOLO FP32 inference speed testing uses inference model exported by `tools/export_model.py` and benchmarked by running `depoly/python/infer.py` with `--run_benchmark`. All testing results 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.
- TensorRT FP16 inference speed testing exclude the time cost of bounding-box decoding(`yolo_box`) part comparing with FP32 testing above, which means that data reading, bounding-box decoding and post-processing(NMS) is excluded(test method same as [YOLOv4(AlexyAB)](https://github.com/AlexeyAB/darknet) too)
- YOLOv4(AlexyAB) performance and inference speed is copy from single Tesla V100 testing results in [YOLOv4 github repo](https://github.com/AlexeyAB/darknet), Tesla V100 TensorRT FP16 inference speed is testing with tkDNN configuration and TensorRT 5.1.2.2 on single Tesla V100 based on [AlexyAB/darknet repo](https://github.com/AlexeyAB/darknet).
-Download and configuration of YOLOv4(AlexyAB) is reproduced model of YOLOv4 in PaddleDetection, whose evaluation performance is same as YOLOv4(AlexyAB), and finetune training is supported in PaddleDetection currently, reproducing by training from backbone pretrain weights is on working, see [PaddleDetection YOLOv4](../yolov4/README.md) for details.
- PP-YOLO tiny is trained on COCO train2017 datast and evaluated on val2017 dataset，`Box AP` is evaluation results as `mAP(IoU=0.5)`.
- PP-YOLO tiny used 4 GPUs for training and mini-batch size as 32 on each GPU, if GPU number and mini-batch size is changed, learning rate and iteration times should be adjusted according [FAQ](../../docs/FAQ.md).
- PP-YOLO tiny inference speeding testing environment and configuration is same as PP-YOLO above.
Training PP-YOLO on 8 GPUs with following command(all commands should be run under PaddleDetection root directory as default), use `--eval` to enable alternate evaluation during training.
For evaluation on COCO test-dev2017 dataset, `configs/ppyolo/ppyolo_test.yml` should be used, please download COCO test-dev2017 dataset from [COCO dataset download](https://cocodataset.org/#download) and decompress to pathes configured by `EvalReader.dataset` in `configs/ppyolo/ppyolo_test.yml` and run evaluation by following command:
```bash
# 使用PaddleDetection发布的权重
# use weights released in PaddleDetection model zoo
Evaluation results will be saved in `bbox.json`, compress it into a `zip` package and upload to [COCO dataset evaluation](https://competitions.codalab.org/competitions/20794#participate) to evaluate.
**NOTE:**`configs/ppyolo/ppyolo_test.yml` is only used for evaluation on COCO test-dev2017 dataset, could not be used for training or COCO val2017 dataset evaluating.
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.
For inference deployment or benchmard, model exported with `tools/export_model.py` should be used and perform inference with Paddle inference library with following commands:
```bash
# 导出模型，默认存储于output/ppyolo目录
# export model, model will be save in output/ppyolo as default
Benchmark testing for PP-YOLO uses model without data reading and post-processing(NMS), export model with `--exclude_nms` to prunce NMS for benchmark testing from mode with following commands:
-Performance and inference spedd are measure with input shape as 608
-All models are trained on COCO train2017 datast and evaluated on val2017 dataset，`Box AP` is evaluation results as `mAP(IoU=0.5:0.95)`.
-Inference speed is tested on single Tesla V100 with batch size as 1 following test method and environment configuration in benchmark above.
-[YOLOv3-DarkNet53](../yolov3_darknet.yml) with mAP as 38.9 is optimized YOLOv3 model in PaddleDetection，see [Model Zoo](../../docs/MODEL_ZOO.md) for details.
