1 / 9

Advanced RGB-D Gesture Recognition with Microsoft Kinect: Innovations and Competition Insights

This project delves into RGB-D gesture recognition utilizing Microsoft Kinect technology, highlighting participation in the ChaLearn competition with a prize pool of up to $100,000. The challenge attracted 53 teams and 572 entries, focused primarily on hand gestures and incorporating head movements and facial expressions. The approach includes one-shot gesture learning, constructing point clouds from multi-view geometry, and implementing a new pose estimation algorithm to enhance joint location accuracy in 3D space. Various innovative methods in action and gesture recognition are explored to improve tracking and segmentation.

sibyl
Télécharger la présentation

Advanced RGB-D Gesture Recognition with Microsoft Kinect: Innovations and Competition Insights

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript

  1. REU Project RGBD gesture recognition with the Microsoft Kinect Steven Hickson

  2. ChaLearn Competition • 20,000$ + up to 100,000$ prize in IP licensing • 53 teams, 80 participants, 572 entries • One shot gesture learning • Can use depth and RGB. • Focused mostly on hand gestures but may include head movements or facial expressions • Rules: http://www.causality.inf.ethz.ch/Gesture/ChaLearnGestureChallengeOfficialRules12-5-11.pdf

  3. Constructing Point Clouds Started constructing point clouds of Chalearn database using multiple view geometry and calibration parameters.

  4. Approach • Implement New Pose Estimation algorithm • Change joint locations to 3D, change Pose Estimation algorithm to take advantage of 3D information. • Enforce rigid joint constraints using proportions between joint distances. • Action recognition, possibly using chaotic invariance.

  5. Old Pose Estimation Results

  6. Efficient Temporal Graph Based Person Segmentation

  7. Skin Chromatic Segmentation and Hand Tracking Takes advantage of human skin color and optical flow

  8. New Pose Estimation Results Error Still Exists

  9. Next: Skeleton Tracking Change SkelTrack Library to work with out input and upper body images

More Related