Repository for Master Thesis:
Reconstructing Human Hands from Egocentric RGB Data @Unipd
Abstract:
Reconstructing 3D poses from a single RGB image is a challenging task. Such computer vision problem hides an inherent ambiguity in the determination of the depth coordinate of the keypoints. In the following work I will start from exploring state-of-the-art approaches used to solve it focusing specifically on Human Hands Pose Estima- tion I will consider the most natural settings, including self-interaction and interaction with objects. Expressing the groundtruth hand label coordinates in the reference frame centered in a standard point (e.g. camera center) or in one of the hand joints, plays an important role to the success of the training process. After evaluating the benefits of choosing a specific one, I propose a multi-stage approach separately regressing root-relative pose and root coordinates in the camera ref- erence frame. Such model is then trained and tested on the novel dataset: H2O dataset (2 Hands and Objects).
Example outputs:
This repository contains modified code taken from: "Interacting Two-Hand 3D Pose and Shape Reconstruction from Single Color Image" (ICCV 2021) "Camera Distance-aware Top-down Approach for 3D Multi-person Pose Estimation from a Single RGB Image" (ICCV 2019)