First-person perspective, right-hand dataset and model for recognizing static hand poses.

[dev-td7]

Currently, the pre-trained model in this repository is capable of recognizing hand poses from a second-person perspective. Meaning, an image taken by you of some other person making a hand pose is the ideal image for recognition.

Need to create a new model and dataset of images of the same hand poses taken from a first-person perspective, i.e. images taken by your camera held by your left hand of your right hand. Refer to the reference image provided in the repository for Indian Sign Language Hand Poses.

You need to create a dataset first. 300 images per label is the ideal number per image. Use the training scripts in the train/ folder to create a new model.

What are hand poses?
Static signs made by your hand - imagine you representing the number '5' from your hand is called a hand pose.