Daniil-Osokin

Hi, I believe this is due to drawing all found keypoints, even with low score. Try to draw keypoints, in which network more or less confident, e.g. which have `>=...

I cannot find the threshold for keypoints in config file. Actually, keypoint confidence is the 3rd value in [`final_results`](https://github.com/HRNet/Higher-HRNet-Human-Pose-Estimation/blob/master/tools/valid.py#L188) (in 3rd dim). Suppose you use `save_valid_image` for visualization, so here...

You can play with parameters for your specific use case. Addition of `person_threshold` is an option, e.g. you can sum all keypoints confidence (`joint[2]` in the code above) and draw...

The tricks above can deal with this. You just need to apply them at the time of poses construction, not at the time of result visualization. Just set higher threshold...

> A's knee was linked to B's ankle This behavior seems buggy for me (but possibly this is a feature to maximize leaderboard score). In general, parts of different persons...

Will [requirements.txt](https://github.com/HRNet/HigherHRNet-Human-Pose-Estimation/blob/master/requirements.txt#L1) work for you?

Hi, I've run [valid.py](https://github.com/HRNet/HigherHRNet-Human-Pose-Estimation/blob/master/tools/valid.py) on COCO validation data.

Actually there are some nice poses in the output. To get rid of mispredicted results a post-processing (which discussed above) is required.

I would say the first thing for this image to get better results is to increase network input resolution (possibly set scale 2 (or even 4) [here](https://github.com/HRNet/HigherHRNet-Human-Pose-Estimation/blob/master/experiments/coco/higher_hrnet/w48_640_adam_lr1e-3.yaml#L109)). Thanks, however this...

You can reproduce this result using `--height-size=1024`.