Skip to content

Hand detection & segmentation using deep neural transfer learning & classical computer vision techniques. - 2022 - University of Padova

Notifications You must be signed in to change notification settings

aonurakman/Human-Hand-Detection

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

7 Commits
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Human Hand Detection & Segmentation

Team members: Onur Akman, Henrik Adrian Hansen, Matthias Schmitz

Phase 1

Detecting bounding boxes surrounding the human hand in the given input image. Used transfer learning with YOLOv3. Fully developed with Python. Output is constructed by fusing the results of multiple weights, and then OpenCV's non-maximum suppression. Each one of the weights used in this step is from the same architecture (YOLOv3) but fine-tuned on a different dataset. This enabled us to use the strengths of each one of them and make up for their weaknesses.

Output of Phase 1

Phase 2

Bounding boxes detected in Phase 1 are the input for Phase 2. This phase is completely developed with C++, using methods built-in OpenCV. The output will be pixel-wise segmentations of the hand regions in the given image. A pipeline is created using multiple techniques, and these are first and second derivative filtering, thresholds, contour finding and filling, segmentation by color (RGB) and the GrabCut algorithm.

Output of Phase 2

Report

More details about the implementation, the results, and the evaluations.

About

Hand detection & segmentation using deep neural transfer learning & classical computer vision techniques. - 2022 - University of Padova

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published