pkd_hands

By David Ong Tat-Wee, Oct 2022

Introduction

This is a set of software to demonstrate how to use MediaPipe's Hands model within PeekingDuck.

The components are:

• src/custom_nodes/dabble/hand_capture.py: PeekingDuck custom node to show live hand and to capture landmark coordinates into SQL database.

• src/custom_nodes/dabble/hand_view.py: PeekingDuck custom node to show rendered hand using data within SQL database.

• src/custom_nodes/model/hand_inference.py: PeekingDuck custom node to perform inference on hand(s) and recognise hand gestures.

• src/hand.py: encapsulates all hand related functions and methods.

• src/train_hand_gestures_classifier.py: script to train a fully connected network for hand gestures classification under Tensorflow.

Run

Run the custom nodes with peekingduck run --config_path <pipeline yaml>

E.g. peekingduck run --config_path inference_hands.yml will run hand inference using PeekingDuck's custom_nodes.dabble.hand_inference node.

Pipeline Files

capture_hands.yml
inference_hands.yml
view_hands.yml

Jira Integration

New note to test Jira integration.