All research areas
Research
Computer Vision
Perception systems that hold up in bad light and worse weather.
Benchmark accuracy is the beginning of the conversation. We care about the long tail: occlusion, motion blur, domain shift, and the edge device that has to run all of it on a power budget.
What this looks like in practice
- Detection, tracking, and pose estimation
- Edge-deployed inference on constrained hardware
- Visual inspection and anomaly detection
- Dataset curation and active learning loops
- Domain adaptation and synthetic data
Related publications
- Designing for the device people actually ownMd Sharif Mollah · 10 July 2026
