Implement a Pure Java AI-Based Palm Tracking Library

Job ID: 39423153

Budget: €30 – €250 EUR

I would like to develop a pure Java software library for AI-based palm tracking, designed to run entirely locally without relying on Python, native libraries (e.g., DLLs or shared objects), or external runtime dependencies. The final application should be distributed as a fat JAR, fully self-contained and ready to run on desktop systems.

Requirements:
1)The implementation must not use traditional techniques such as Haar Cascade Classifiers.
2)It is acceptable to use Java-based AI libraries, such as Deep Java Library (DJL){https://djl.ai/}, as long as they do not depend on native code.
3)The model should ideally be based on or trained using the EgoHands dataset{https://vision.soic.indiana.edu/projects/egohands/}, or compatible with it.
4)The library must detect palms in the input image and return their bounding box coordinates.
5)Detection of fingers or hand joints is considered a plus, but not strictly required.
6)The software is intended for desktop environments only (no mobile or embedded support required).

Output:
For each detected palm, the system should return the bounding box coordinates within the image.
Optionally, it may also return keypoint locations for finger joints if the model supports it.