Closed Mayo001 closed 11 months ago
Hello, thanks for your comment. Since the model predicts 2.5d coordinates, the xy coordinates in the model predictions are just the joint coordinates in the image coordinate system that you need. Therefore, you will find that when you remove the few lines of coordinate conversion code after #visualization, the visualized results are normal. As for those conversion codes, they are to convert the 2.5d coordinates in the image coordinate system into 3d camera coordinate system. If you don't understand 2.5d coordinates, you can read the article that proposed the InterHand2.6m dataset. All coordinate transformation methods in our model are the same as the baseline in that article. Hope this can answer your questions!
Thank u very much! I got the desired result after removing the coordinate conversion code.
Thanks for your code! I found that the estimated key points do not correspond to ground-truth when I visualize it directly from preds. What should I do to recover the estimated 3D coordinates to a 2D image. btw, I'm a beginner, please don't hesitate to teach!