TexasInstruments / edgeai-yolov5

YOLOv5 🚀 in PyTorch > ONNX > CoreML > TFLite. Forked from https://ultralytics.com/yolov5
https://github.com/TexasInstruments/edgeai
GNU General Public License v3.0
640 stars 121 forks source link

This points is not correct? #56

Closed CYYAI closed 2 years ago

CYYAI commented 2 years ago

image

debapriyamaji commented 2 years ago

@cyymt , Which model did you try? Did you take the latest models and try it out again? E.g. http://software-dl.ti.com/jacinto7/esd/modelzoo/gplv3/08_02_00_11/edgeai-yolov5/pretrained_models/checkpoints/keypoint/coco/edgeai-yolov5/other/best_models/yolov5s6_640_60p7_85p3_kpts_head_6x_dwconv_3x3_lr_0p01/weights/last.pt

I have one request: Will you please try inference on the *ti-lite models and let me know if you see any issue with them. https://github.com/TexasInstruments/edgeai-yolov5/tree/yolo-pose#yolov5-ti-lite-based-models-and-ckpts

Regards, Debapriya

wanglegoc commented 2 years ago

hi @cyymt do you know why one of the keypoint isn't right? I obtain some results like yours. Thanks

debapriyamaji commented 2 years ago

@wanglegoc @cyymt Please try the ti-lite model as suggested above. https://github.com/TexasInstruments/edgeai-yolov5/tree/yolo-pose#yolov5-ti-lite-based-models-and-ckpts

Can you also share the exact command that you used?