Skip to content

aalpatya/detect_hands

Repository files navigation

SSD on CPU using subset of Egohands Dataset

This is a hands detector that uses the SSD_mobilenet_v2_fpn_320 network. It has been trained with transfer learning, on a small subset (about 300 images) from the EgoHands dataset: http://vision.soic.indiana.edu/projects/egohands/

The dataset has been cleaned (sometimes bounding boxes in images had x_min = x_max, and y_min = y_max). Tutorial: https://aalpatya.medium.com/train-an-object-detector-using-tensorflow-2-object-detection-api-in-2021-a4fed450d1b9

Acknowledgements:

Requires tensorflow 2, (tested on tensorflow 2.4.1) and OpenCV 4 (tested on OpenCV 4.2.1)

Live Hand Detection only

python3 webcam_detect_hands.py

out1

Live Hand Theremin

python3 theremin_hands.py

See it in action by clicking this thumbnail:

Hand Theremin

(https://www.youtube.com/watch?v=3Kw0j-96lWc)

Live Mouse Theremin

python3 theremin_mouse.py

TODO: Put camera frame reading and processing into multi-threading

About

SSD on CPU using subset of Egohands Dataset

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published