Lightweight human detection models generated on high-quality human data sets. It can detect objects with high accuracy and speed in a total of 8 classes: whole body, whole body with wheelchair, head, face, hands, left hand, right hand, and foot(feet). Even the classification problem is being attempted to be solved by object detection. There is no need to perform any complex affine transformations or other processing for pre-processing and post-processing of input images. In addition, the resistance to Motion Blur, Gaussian noise, contrast noise, backlighting, and halation is quite strong because it was trained only on images with added photometric noise for all images in the MS-COCO subset of the image set. In addition, about half of the image set was annotated by me with the aspect ratio of the original image substantially destroyed. I manually annotated all images in the dataset by myself. The model is intended to use real-world video for inference and has enhanced resistance to all kinds of noise. Probably stronger than any known model. However, the quality of the known data set and my data set are so different that an accurate comparison of accuracy is not possible.
Don't be ruled by the curse of mAP.
The use of CD-COCO: Complex Distorted COCO database for Scene-Context-Aware computer vision has also greatly improved resistance to various types of noise.
-
Global distortions
- Noise
- Contrast
- Compression
- Photorealistic Rain
- Photorealistic Haze
- Motion-Blur
- Defocus-Blur
- Backlight illumination
-
Local distortions
- Motion-Blur
- Defocus-Blur
- Backlight illumination
-
Highly accurate detection results
I don't evaluate by Cherry-picked data, Best-case data or Biased data at all. Therefore, only difficult images and situations in which the model is most prone to detection errors are used for validation.
-
Real-time - YOLOv9-E End-to-End (Pre-process/Post-Process) ONNX + TensorRT + USB Camera
output_e_swish.mp4
-
Still images
-
- COCO-Hand http://vision.cs.stonybrook.edu/~supreeth/COCO-Hand.zip
- CD-COCO: Complex Distorted COCO database for Scene-Context-Aware computer vision
- I am adding my own enhancement data to COCO-Hand and re-annotating all images. In other words, only COCO images were cited and no annotation data were cited.
- I have no plans to publish my own dataset.
- Annotation quantity
TOTAL: 10,578 images TOTAL: 254,459 labels train - 201,879 labels { "body": 49,413, "body_with_wheelchair": 580, "head": 42,155, "face": 22,680, "hand": 30,481, "hand_left": 15,257, "hand_right": 15,223, "foot": 26,090 } val - 52,580 labels { "body": 13,119, "body_with_wheelchair": 150, "head": 10,839, "face": 5,953, "hand": 7,851, "hand_left": 3,921, "hand_right": 3,929, "foot": 6,818 }
Halfway compromises are never acceptable.
| Class Name | Class ID |
|---|---|
| Body | 0 |
| Body-with-Wheelchair | 1 |
| Head | 2 |
| Face | 3 |
| Hand | 4 |
| Left-Hand | 5 |
| Right-Hand | 6 |
| Foot (Feet) | 7 |
-
Python 3.10
-
onnx 1.14.1+
-
onnxruntime-gpu v1.16.1 (TensorRT Execution Provider Enabled Binary. See: onnxruntime-gpu v1.16.1 + CUDA 11.8 + TensorRT 8.5.3 build (RTX3070))
-
opencv-contrib-python 4.9.0.80+
-
numpy 1.24.3
-
TensorRT 8.5.3-1+cuda11.8
# Common ############################################ pip install opencv-contrib-python numpy onnx # For ONNX ########################################## pip uninstall onnxruntime onnxruntime-gpu pip install onnxruntime or pip install onnxruntime-gpu
-
Demonstration of models with built-in post-processing (Float32/Float16)
usage: demo_yolov9_onnx_handLR_foot_wheelchair.py \ [-h] \ [-m MODEL] \ (-v VIDEO | -i IMAGES_DIR) \ [-ep {cpu,cuda,tensorrt}] \ [-it] \ [-dvw] \ [-dwk] \ [-dlr] options: -h, --help show this help message and exit -m MODEL, --model MODEL ONNX/TFLite file path for YOLOv9. -v VIDEO, --video VIDEO Video file path or camera index. -i IMAGES_DIR, --images_dir IMAGES_DIR jpg, png images folder path. -ep {cpu,cuda,tensorrt}, \ --execution_provider {cpu,cuda,tensorrt} Execution provider for ONNXRuntime. -it {fp16,int8}, --inference_type {fp16,int8} Inference type. Default: fp16 -dvw, --disable_video_writer Disable video writer. Eliminates the file I/O load associated with automatic recording to MP4. Devices that use a MicroSD card or similar for main storage can speed up overall processing. -dwk, --disable_waitKey Disable cv2.waitKey(). When you want to process a batch of still images, disable key-input wait and process them continuously. -dlr, --disable_left_and_right_hand_discrimination_mode Disable left and right hand discrimination mode. -
YOLOv9-Wholebody-with-Wheelchair - N - Swish/SiLU (PINTO original implementation, 2.4 MB)
Class Images Instances mAP50-95 YOLOX-N YOLOv9-N all 2117 48988 0.353 0.352 Body 2117 11624 0.392 0.444 BodyWithWheelchair 2117 153 0.587 0.571 Head 2117 9936 0.467 0.516 Face 2117 5653 0.362 0.397 Hand 2117 7525 0.306 0.266 Hand-Left 2117 3739 0.237 0.198 Hand-Right 2117 3786 0.241 0.198 Foot 2117 6572 0.231 0.225 -
YOLOv9-Wholebody-with-Wheelchair - N - ReLU (PINTO original implementation, 2.4 MB, For INT8/QAT)
Class Images Instances mAP50-95 YOLOX-N YOLOv9-N all 2117 48988 0.353 0.351 Body 2117 11624 0.392 0.435 BodyWithWheelchair 2117 153 0.587 0.590 Head 2117 9936 0.467 0.505 Face 2117 5653 0.362 0.397 Hand 2117 7525 0.306 0.257 Hand-Left 2117 3739 0.237 0.200 Hand-Right 2117 3786 0.241 0.201 Foot 2117 6572 0.231 0.222 -
YOLOv9-Wholebody-with-Wheelchair - T - Swish/SiLU
Class Images Instances mAP50-95 YOLOX-T YOLOv9-T all 2117 48988 0.421 0.477 Body 2117 11624 0.463 0.590 BodyWithWheelchair 2117 153 0.674 0.771 Head 2117 9936 0.507 0.591 Face 2117 5653 0.417 0.468 Hand 2117 7525 0.372 0.401 Hand-Left 2117 3739 0.325 0.335 Hand-Right 2117 3786 0.318 0.327 Foot 2117 6572 0.291 0.332 -
YOLOv9-Wholebody-with-Wheelchair - T - ReLU
Class Images Instances mAP50-95 YOLOX-T YOLOv9-T all 2117 48988 0.421 0.471 Body 2117 11624 0.463 0.567 BodyWithWheelchair 2117 153 0.674 0.764 Head 2117 9936 0.507 0.584 Face 2117 5653 0.417 0.486 Hand 2117 7525 0.372 0.390 Hand-Left 2117 3739 0.325 0.333 Hand-Right 2117 3786 0.318 0.323 Foot 2117 6572 0.291 0.320 -
YOLOv9-Wholebody-with-Wheelchair - S - Swish/SiLU
Class Images Instances mAP50-95 YOLOX-X YOLOv9-S all 2117 48988 0.554 0.560 Body 2117 11624 0.614 0.672 BodyWithWheelchair 2117 153 0.871 0.844 Head 2117 9936 0.585 0.646 Face 2117 5653 0.506 0.555 Hand 2117 7525 0.513 0.486 Hand-Left 2117 3739 0.456 0.432 Hand-Right 2117 3786 0.449 0.431 Foot 2117 6572 0.431 0.411 -
YOLOv9-Wholebody-with-Wheelchair - S - ReLU
Class Images Instances mAP50-95 YOLOX-X YOLOv9-S all 2117 48988 0.554 0.556 Body 2117 11624 0.614 0.659 BodyWithWheelchair 2117 153 0.871 0.835 Head 2117 9936 0.585 0.640 Face 2117 5653 0.506 0.561 Hand 2117 7525 0.513 0.480 Hand-Left 2117 3739 0.456 0.430 Hand-Right 2117 3786 0.449 0.436 Foot 2117 6572 0.431 0.404 -
YOLOv9-Wholebody-with-Wheelchair - M - Swish/SiLU
Class Images Instances mAP50-95 YOLOX-X YOLOv9-M all 2117 48988 0.554 0.577 Body 2117 11624 0.614 0.668 BodyWithWheelchair 2117 153 0.871 0.814 Head 2117 9936 0.585 0.666 Face 2117 5653 0.506 0.601 Hand 2117 7525 0.513 0.509 Hand-Left 2117 3739 0.456 0.475 Hand-Right 2117 3786 0.449 0.467 Foot 2117 6572 0.431 0.417 -
YOLOv9-Wholebody-with-Wheelchair - M - ReLU
Class Images Instances mAP50-95 YOLOX-X YOLOv9-M all 2117 48988 0.554 0.575 Body 2117 11624 0.614 0.663 BodyWithWheelchair 2117 153 0.871 0.788 Head 2117 9936 0.585 0.664 Face 2117 5653 0.506 0.605 Hand 2117 7525 0.513 0.513 Hand-Left 2117 3739 0.456 0.477 Hand-Right 2117 3786 0.449 0.475 Foot 2117 6572 0.431 0.414 -
YOLOv9-Wholebody-with-Wheelchair - C - Swish/SiLU
Class Images Instances mAP50-95 YOLOX-X YOLOv9-C all 2117 48988 0.554 0.592 Body 2117 11624 0.614 0.687 BodyWithWheelchair 2117 153 0.871 0.822 Head 2117 9936 0.585 0.674 Face 2117 5653 0.506 0.608 Hand 2117 7525 0.513 0.527 Hand-Left 2117 3739 0.456 0.491 Hand-Right 2117 3786 0.449 0.489 Foot 2117 6572 0.431 0.436 -
YOLOv9-Wholebody-with-Wheelchair - C - ReLU
Class Images Instances mAP50-95 YOLOX-X YOLOv9-C all 2117 48988 0.554 0.593 Body 2117 11624 0.614 0.679 BodyWithWheelchair 2117 153 0.871 0.825 Head 2117 9936 0.585 0.676 Face 2117 5653 0.506 0.617 Hand 2117 7525 0.513 0.529 Hand-Left 2117 3739 0.456 0.491 Hand-Right 2117 3786 0.449 0.491 Foot 2117 6572 0.431 0.437 -
YOLOv9-Wholebody-with-Wheelchair - E - Swish/SiLU
Class Images Instances mAP50-95 YOLOX-X YOLOv9-E all 2117 48988 0.554 0.660 Body 2117 11624 0.614 0.748 BodyWithWheelchair 2117 153 0.871 0.865 Head 2117 9936 0.585 0.723 Face 2117 5653 0.506 0.675 Hand 2117 7525 0.513 0.606 Hand-Left 2117 3739 0.456 0.571 Hand-Right 2117 3786 0.449 0.573 Foot 2117 6572 0.431 0.523 -
YOLOv9-Wholebody-with-Wheelchair - E - ReLU
Class Images Instances mAP50-95 YOLOX-X YOLOv9-E all 2117 48988 0.554 0.647 Body 2117 11624 0.614 0.731 BodyWithWheelchair 2117 153 0.871 0.850 Head 2117 9936 0.585 0.719 Face 2117 5653 0.506 0.673 Hand 2117 7525 0.513 0.589 Hand-Left 2117 3739 0.456 0.556 Hand-Right 2117 3786 0.449 0.552 Foot 2117 6572 0.431 0.508 -
Pre-Process
To ensure fair benchmark comparisons with YOLOX,
BGR to RGB conversion processingandnormalization by division by 255.0are added to the model input section. In addition, aresizing processfor input images has been added to improve operational flexibility. Thus, in any model, inferences can be made at any image size. The string1x3x{H}x{W}at the end of the file name does not indicate the input size of the image, but the processing resolution inside the model. Therefore, the smaller the values of{H}and{W}, the lower the computational cost and the faster the inference speed. Models with larger values of{H}and{W}increase the computational cost and decrease the inference speed. Since the concept is different from the resolution of an image, any size image can be batch processed. e.g. 240x320, 480x640, 720x1280, ...
-
Post-Process
Because I add my own post-processing to the end of the model, which can be inferred by TensorRT, CUDA, and CPU, the benchmarked inference speed is the end-to-end processing speed including all pre-processing and post-processing. EfficientNMS in TensorRT is very slow and should be offloaded to the CPU.
-
NMS default parameter
param value note max_output_boxes_per_class 20 Maximum number of outputs per class of one type. 20indicates that the maximum number of people detected is20, the maximum number of heads detected is20, and the maximum number of hands detected is20. The larger the number, the more people can be detected, but the inference speed slows down slightly due to the larger overhead of NMS processing by the CPU. In addition, as the number of elements in the final output tensor increases, the amount of information transferred between hardware increases, resulting in higher transfer costs on the hardware circuit. Therefore, it would be desirable to set the numerical size to the minimum necessary.iou_threshold 0.40 A value indicating the percentage of occlusion allowed for multiple bounding boxes of the same class. 0.40is excluded from the detection results if, for example, two bounding boxes overlap in more than 41% of the area. The smaller the value, the more occlusion is tolerated, but over-detection may increase.score_threshold 0.25 Bounding box confidence threshold. Specify in the range of 0.00to1.00. The larger the value, the stricter the filtering and the lower the NMS processing load, but in exchange, all but bounding boxes with high confidence values are excluded from detection. This is a parameter that has a very large percentage impact on NMS overhead. -
Change NMS parameters
Use PINTO0309/sam4onnx to rewrite the
NonMaxSuppressionparameter in the ONNX file.For example,
pip install onnxsim==0.4.33 \ && pip install -U simple-onnx-processing-tools \ && pip install -U onnx \ && python -m pip install -U onnx_graphsurgeon \ --index-url https://pypi.ngc.nvidia.com ### max_output_boxes_per_class ### Example of changing the maximum number of detections per class to 100. sam4onnx \ --op_name main01_nonmaxsuppression11 \ --input_onnx_file_path yolov9_s_wholebody_with_wheelchair_post_0100_1x3x256x320.onnx \ --output_onnx_file_path yolov9_s_wholebody_with_wheelchair_post_0100_1x3x256x320.onnx \ --input_constants main01_max_output_boxes_per_class int64 [100] ### iou_threshold ### Example of changing the allowable area of occlusion to 20%. sam4onnx \ --op_name main01_nonmaxsuppression11 \ --input_onnx_file_path yolov9_s_wholebody_with_wheelchair_post_0100_1x3x256x320.onnx \ --output_onnx_file_path yolov9_s_wholebody_with_wheelchair_post_0100_1x3x256x320.onnx \ --input_constants main01_iou_threshold float32 [0.20] ### score_threshold ### Example of changing the bounding box score threshold to 15%. sam4onnx \ --op_name main01_nonmaxsuppression11 \ --input_onnx_file_path yolov9_s_wholebody_with_wheelchair_post_0100_1x3x256x320.onnx \ --output_onnx_file_path yolov9_s_wholebody_with_wheelchair_post_0100_1x3x256x320.onnx \ --input_constants main01_score_threshold float32 [0.15]
-
Post-processing structure
PyTorch alone cannot generate this post-processing. For operational flexibility,
EfficientNMSis not used.
-
-
INT8 quantization (YOLOv9-QAT)
If this work has contributed in any way to your research or business, I would be happy to be cited in your literature.
@software{YOLOv9-Wholebody-with-Wheelchair,
author={Katsuya Hyodo},
title={Lightweight human detection models generated on high-quality human data sets. It can detect objects with high accuracy and speed in a total of eight classes: whole body, whole body with wheelchair, head, face, hands, left hand, right hand, and foot(feet).},
url={https://github.com/PINTO0309/PINTO_model_zoo/tree/main/450_YOLOv9-Wholebody-with-Wheelchair},
year={2024},
month={6},
doi={10.5281/zenodo.10229410}
}I am very grateful for their excellent work.
-
COCO-Hand
https://vision.cs.stonybrook.edu/~supreeth/
@article{Hand-CNN, title={Contextual Attention for Hand Detection in the Wild}, author={Supreeth Narasimhaswamy and Zhengwei Wei and Yang Wang and Justin Zhang and Minh Hoai}, booktitle={International Conference on Computer Vision (ICCV)}, year={2019}, url={https://arxiv.org/pdf/1904.04882.pdf} }
-
CD-COCO: Complex Distorted COCO database for Scene-Context-Aware computer vision
@INPROCEEDINGS{10323035, author={Beghdadi, Ayman and Beghdadi, Azeddine and Mallem, Malik and Beji, Lotfi and Cheikh, Faouzi Alaya}, booktitle={2023 11th European Workshop on Visual Information Processing (EUVIP)}, title={CD-COCO: A Versatile Complex Distorted COCO Database for Scene-Context-Aware Computer Vision}, year={2023}, volume={}, number={}, pages={1-6}, doi={10.1109/EUVIP58404.2023.10323035} }
-
YOLOv9
https://github.com/WongKinYiu/yolov9
@article{wang2024yolov9, title={{YOLOv9}: Learning What You Want to Learn Using Programmable Gradient Information}, author={Wang, Chien-Yao and Liao, Hong-Yuan Mark}, booktitle={arXiv preprint arXiv:2402.13616}, year={2024} }
-
YOLOv9-QAT