Human-Following Bot for the Indian Army
Deep Learning · Generative Adversarial Networks · Sept 2023 – May 2024 · Ladakh, India
A vision-guided robot designed to follow a person reliably even in heavy rain and at night — conditions where standard perception pipelines fail.
- Built GAN-based data augmentation and feature-preservation pipelines so the robot can “see” in rainy, low-light conditions.
- Implemented a synthetic rain generator (RainGAN) and CycleGAN for cross-domain image translation.
- Achieved an FID score of 31.4 on generated scenes.
- Selected and funded under NewGenIEDC-IIITA and chosen by the Indian Army.
Related publications: RainGAN-Kathmandu (IEEE TENCON 2025) and ADAS Framework Utilizing CycleGAN (CONFLUENCE 2025).
