This repository tracks the progress of tasks and experiments carried out as part of the CDAC Internship.
All the demo videos for the project tasks are available here:
π Simulation Demo Folder (Google Drive):
https://drive.google.com/drive/folders/1f5XfqfWDF59QwHiE3CUAL9E3sC4UmrPh
Task 1: Basic Navigation Stack (Completed βοΈ)
- Implemented and tested basic navigation and control stack in simulation.
Task 2: Visual Semantic SLAM Demo (Completed βοΈ)
- Built a demo pipeline for visual semantic SLAM in a simulated environment.
Task 3: ZED2i Camera Exploration (Completed βοΈ)
- Studied ZED2i camera capabilities.
- Explored possible perception pipelines using only ZED2i (RGB + depth) for navigation and tracking.
Task 4: Tracking Continuity (Occlusion Handling)
- AirSim + Python-based tracking with movement continuity in simulation.
- Achieved continuous tracking despite temporary crowd occlusion.
Task 5: Re-identification After Temporary Loss
- Implemented re-identification logic when the target is temporarily lost.
- Drone performs localized search and resumes tracking on reappearance.
Task 6: Colour of Clothes Identification (Attribute Extraction)
- Designed colour-based attribute extraction for the target.
- Implemented dress colour tagging (e.g., identifying a child by clothing colour).
Task 7: Appearance-based Re-identification (Limitation)
- Attribute-based tracking validated in simulation.
- Re-identification using clothing attributes: Completed βοΈ
- Re-identification using full appearance features: Limited by simulation realism (ongoing / future work).