Peng Lu

This bug is caused by the distinct format of metainfo in MMPose and MMDetection. Thank you very much for pointing it out. You can modify your code as in https://github.com/open-mmlab/mmpose/pull/1813...

Hi, thanks for using MMPose. If you want to save model outputs during testing, the file path you specify using argument `--out` should have postfix `.json`, `yml`, or `pkl`.

You could try the command `python tools\test.py configs\pig\pig_behavior.py work_dirs\pig_behavior\best_PCK_epoch_80.pth --out data\pigBehavior\results.json` and check if the output file `data\pigBehavior\results.json` contains predicted keypoints.

Hi, if your target objects are included in [COCO category](https://gist.github.com/AruniRC/7b3dadd004da04c80198557db5da4bda), you can directly use detectors trained on COCO such as the one in the following command ```bash python demo/top_down_img_demo_with_mmdet.py \...

Many popular pose estimation benchmarks do not include cross-validation in their protocols. You may need to manually split train/val set in dataset annotation files for cross-validation.

Sorry, there is a bug in the `AdaptiveWingLoss` class. The `output` and `target_weights` have different numbers of dimensions, so they cannot be multiplied directly. You can fix this bug by...

Thanks for bringing this problem to our attention. We will investigate it soon. As you are using only one dataset for evaluation, you can try to set the content in...

貌似其中一张卡有其他进程占用显存导致剩余显存不足？ 下面这个社区群应该还能直接扫码进群 

MMDeploy提供了RTMPose rknn部署的config https://github.com/open-mmlab/mmdeploy/blob/main/configs/mmpose/pose-detection_simcc_rknn-fp16_static-256x192.py 把部署教程里面onnx/tensorrt config换成对应的rknn config就行了

How about supporting [UniFormer](https://github.com/Sense-X/UniFormer/tree/main/pose_estimation)?