idea-research/click-pose

[ICCV 2023] Official implementation of the paper "Neural Interactive Keypoint Detection"

click-pose's Issues

About "First" interactive keypoint detection.

Thank you for your diligent work and valuable contribution to the field.

I would like to point out a potential oversight regarding the claim of originality in your paper.
Specifically, the statement

"we define a novel task called interactive keypoint detection. It aims to effectively maximize benefits of the model to minimize manual effort, and mitigate unfriendly consequences of model failures in out-of-distribution and new-task annotations that increase the need for manual intervention. Accordingly, we present the first neural interactive keypoint detection framework, ClickPose, as a baseline for further research. "

seems to overlook the prior work titled "Morphology-Aware Interactive Keypoint Estimation," presented at MICCAI 22 (https://arxiv.org/pdf/2209.07163.pdf).

Given that this earlier work is referenced in your paper, I believe there might be an awareness of its existence.
To avoid any misunderstanding or misrepresentation, could you please provide some clarification on this matter?

Thank you for considering my inquiry, and I look forward to your response.

Not able to change visibility of keypoints in quick start

Wen I use the skeleton annotator to annotate keypoints, I am not able to change the visibility of keypoints. And the bounding box that comes along with the keypoints in the skeleton tool is not getting stored in the json file when I export the annotation. How do I rectify these issues?