Lightweight human detection models generated on high-quality human data sets. It can detect objects with high accuracy and speed in a total of seven classes: whole body, 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.
Objects up to a minimum size of
4x3pixels can be detected. However, it is difficult to classify left and right hands for an object of that size and is classified asUnknown. The detection of the male head in the center somehow fails. Failure to detect a woman in white on the dog's head.
Strong detection even when feet are heavily shielded by grass.
It strongly detects even if Gaussian noise is added with an intensity that would not be possible in the real world.
Simultaneous acquisition of hand and foot context has greatly improved the performance of distinguishing between limbs.
-
Demonstration of detection of feet wearing black socks and bare feet in all-black, hard-to-define clothing
output_x_foot.mp4
- 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: 10,064 images TOTAL: 244,388 labels train - 193,419 labels { "body": 47,985, "head": 40,422, "face": 21,800, "hand": 29,150, "hand_left": 14,608, "hand_right": 14,541, "foot": 24,913 } val - 50,969 labels { "body": 12,831, "head": 11,006, "face": 5,771, "hand": 7,549, "hand_left": 3,790, "hand_right": 3,758, "foot": 6,264 }
Halfway compromises are never acceptable.
| Class Name | Class ID |
|---|---|
| Body | 0 |
| Head | 1 |
| Face | 2 |
| Hand | 3 |
| Left-Hand | 4 |
| Right-Hand | 5 |
| Foot (Feet) | 6 |
-
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_yolox_onnx_handLR_foot.py \ [-h] \ [-m MODEL] \ (-v VIDEO | -i IMAGES_DIR) \ [-ep {cpu,cuda,tensorrt}] \ [-dvw] \ [-dwk] options: -h, --help show this help message and exit -m MODEL, --model MODEL ONNX/TFLite file path for YOLOX. -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. -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. -
YOLOX-Body-Head-Face-HandLR-Foot-Dist - Nano
Average Precision (AP) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.324 Average Precision (AP) @[ IoU=0.50 | area= all | maxDets=100 ] = 0.617 Average Precision (AP) @[ IoU=0.75 | area= all | maxDets=100 ] = 0.300 Average Precision (AP) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.218 Average Precision (AP) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.501 Average Precision (AP) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.623 Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 1 ] = 0.149 Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 10 ] = 0.355 Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.425 Average Recall (AR) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.327 Average Recall (AR) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.614 Average Recall (AR) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.698 per class AP: | class | AP | class | AP | class | AP | |:--------|:-------|:----------|:-------|:-----------|:-------| | body | 39.832 | head | 46.236 | face | 38.059 | | hand | 30.681 | hand_left | 23.710 | hand_right | 24.494 | | foot | 23.959 | | | | | per class AR: | class | AR | class | AR | class | AR | |:--------|:-------|:----------|:-------|:-----------|:-------| | body | 48.183 | head | 51.642 | face | 44.205 | | hand | 41.074 | hand_left | 37.824 | hand_right | 38.492 | | foot | 36.420 | | | | | -
YOLOX-Body-Head-Face-HandLR-Foot-Dist - Tiny
Average Precision (AP) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.394 Average Precision (AP) @[ IoU=0.50 | area= all | maxDets=100 ] = 0.708 Average Precision (AP) @[ IoU=0.75 | area= all | maxDets=100 ] = 0.388 Average Precision (AP) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.281 Average Precision (AP) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.585 Average Precision (AP) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.684 Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 1 ] = 0.172 Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 10 ] = 0.404 Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.473 Average Recall (AR) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.375 Average Recall (AR) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.661 Average Recall (AR) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.739 per class AP: | class | AP | class | AP | class | AP | |:--------|:-------|:----------|:-------|:-----------|:-------| | body | 46.426 | head | 50.349 | face | 42.593 | | hand | 38.301 | hand_left | 33.715 | hand_right | 33.745 | | foot | 30.351 | | | | | per class AR: | class | AR | class | AR | class | AR | |:--------|:-------|:----------|:-------|:-----------|:-------| | body | 53.056 | head | 54.960 | face | 48.086 | | hand | 46.050 | hand_left | 44.556 | hand_right | 44.685 | | foot | 40.043 | | | | | -
YOLOX-Body-Head-Face-HandLR-Foot-Dist - S
Average Precision (AP) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.420 Average Precision (AP) @[ IoU=0.50 | area= all | maxDets=100 ] = 0.716 Average Precision (AP) @[ IoU=0.75 | area= all | maxDets=100 ] = 0.425 Average Precision (AP) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.295 Average Precision (AP) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.631 Average Precision (AP) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.756 Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 1 ] = 0.177 Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 10 ] = 0.435 Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.508 Average Recall (AR) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.401 Average Recall (AR) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.712 Average Recall (AR) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.816 per class AP: | class | AP | class | AP | class | AP | |:--------|:-------|:----------|:-------|:-----------|:-------| | body | 52.208 | head | 52.599 | face | 45.812 | | hand | 43.065 | hand_left | 32.875 | hand_right | 32.383 | | foot | 34.901 | | | | | per class AR: | class | AR | class | AR | class | AR | |:--------|:-------|:----------|:-------|:-----------|:-------| | body | 58.129 | head | 57.021 | face | 50.979 | | hand | 49.781 | hand_left | 48.122 | hand_right | 47.827 | | foot | 43.907 | | | | | -
YOLOX-Body-Head-Face-HandLR-Foot-Dist - M
Average Precision (AP) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.473 Average Precision (AP) @[ IoU=0.50 | area= all | maxDets=100 ] = 0.774 Average Precision (AP) @[ IoU=0.75 | area= all | maxDets=100 ] = 0.488 Average Precision (AP) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.343 Average Precision (AP) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.690 Average Precision (AP) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.834 Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 1 ] = 0.195 Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 10 ] = 0.471 Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.543 Average Recall (AR) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.435 Average Recall (AR) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.749 Average Recall (AR) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.867 per class AP: | class | AP | class | AP | class | AP | |:--------|:-------|:----------|:-------|:-----------|:-------| | body | 57.555 | head | 54.713 | face | 48.840 | | hand | 48.599 | hand_left | 41.270 | hand_right | 40.439 | | foot | 39.804 | | | | | per class AR: | class | AR | class | AR | class | AR | |:--------|:-------|:----------|:-------|:-----------|:-------| | body | 62.326 | head | 58.714 | face | 53.595 | | hand | 53.605 | hand_left | 52.609 | hand_right | 52.102 | | foot | 47.186 | | | | | -
YOLOX-Body-Head-Face-HandLR-Foot-Dist - L
Average Precision (AP) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.498 Average Precision (AP) @[ IoU=0.50 | area= all | maxDets=100 ] = 0.798 Average Precision (AP) @[ IoU=0.75 | area= all | maxDets=100 ] = 0.520 Average Precision (AP) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.369 Average Precision (AP) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.715 Average Precision (AP) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.857 Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 1 ] = 0.201 Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 10 ] = 0.487 Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.561 Average Recall (AR) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.453 Average Recall (AR) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.768 Average Recall (AR) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.892 per class AP: | class | AP | class | AP | class | AP | |:--------|:-------|:----------|:-------|:-----------|:-------| | body | 60.147 | head | 55.945 | face | 50.340 | | hand | 50.840 | hand_left | 44.890 | hand_right | 44.216 | | foot | 42.438 | | | | | per class AR: | class | AR | class | AR | class | AR | |:--------|:-------|:----------|:-------|:-----------|:-------| | body | 64.474 | head | 59.821 | face | 55.076 | | hand | 55.525 | hand_left | 54.706 | hand_right | 54.475 | | foot | 48.955 | | | | | -
YOLOX-Body-Head-Face-HandLR-Foot-Dist - X
Average Precision (AP) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.524 Average Precision (AP) @[ IoU=0.50 | area= all | maxDets=100 ] = 0.821 Average Precision (AP) @[ IoU=0.75 | area= all | maxDets=100 ] = 0.554 Average Precision (AP) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.397 Average Precision (AP) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.738 Average Precision (AP) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.869 Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 1 ] = 0.209 Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 10 ] = 0.506 Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.579 Average Recall (AR) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.471 Average Recall (AR) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.784 Average Recall (AR) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.897 per class AP: | class | AP | class | AP | class | AP | |:--------|:-------|:----------|:-------|:-----------|:-------| | body | 61.893 | head | 57.409 | face | 52.849 | | hand | 53.589 | hand_left | 48.605 | hand_right | 47.634 | | foot | 44.999 | | | | | per class AR: | class | AR | class | AR | class | AR | |:--------|:-------|:----------|:-------|:-----------|:-------| | body | 65.588 | head | 60.996 | face | 57.313 | | hand | 57.589 | hand_left | 57.127 | hand_right | 56.292 | | foot | 50.565 | | | | | -
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 yolox_s_body_head_hand_post_0299_0.4983_1x3x256x320.onnx \ --output_onnx_file_path yolox_s_body_head_hand_post_0299_0.4983_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 yolox_s_body_head_hand_post_0299_0.4983_1x3x256x320.onnx \ --output_onnx_file_path yolox_s_body_head_hand_post_0299_0.4983_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 yolox_s_body_head_hand_post_0299_0.4983_1x3x256x320.onnx \ --output_onnx_file_path yolox_s_body_head_hand_post_0299_0.4983_1x3x256x320.onnx \ --input_constants main01_score_threshold float32 [0.15]
-
Post-processing structure
PyTorch alone cannot generate this post-processing.
-
-
INT8 quantization (TexasInstruments/YOLOX-ti-lite)
In my experience, YOLOX has a very large accuracy degradation during quantization due to its structure. The reasons for this and the workaround are examined in detail by TexasInstruments. I have summarized the main points below on how to minimize accuracy degradation during quantization through my own practice. I just put into practice what TexasInstruments suggested, but the degrade in accuracy during quantization was extremely small. Note, however, that the results of the Float16 mixed-precision training before quantization are significantly degraded in accuracy due to the change in activation function to
ReLUand many other workarounds, as well as the completely different data sets being benchmarked.
If this work has contributed in any way to your research or business, I would be happy to be cited in your literature.
@software{YOLOX-Body-Head-Face-HandLR-Foot-Dist,
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 seven classes: whole body, head, face, hands, left hand, right hand, and foot(feet).},
url={https://github.com/PINTO0309/PINTO_model_zoo/tree/main/445_YOLOX-Body-Head-Face-HandLR-Foot-Dist},
year={2024},
month={2},
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} }
-
YOLOX
https://github.com/Megvii-BaseDetection/YOLOX
@article{yolox2021, title={YOLOX: Exceeding YOLO Series in 2021}, author={Ge, Zheng and Liu, Songtao and Wang, Feng and Li, Zeming and Sun, Jian}, journal={arXiv preprint arXiv:2107.08430}, year={2021} }
-
YOLOX-ti-lite
-
yolox-ti-lite_tflite
-
YOLOX-Colaboratory-Training-Sample
高橋かずひと https://github.com/Kazuhito00
https://github.com/Kazuhito00/YOLOX-Colaboratory-Training-Sample