Perception for the streets nobody designed a dataset for.
Lynx is a robotics and perception research group at BRAC University, building the vision, sensing, and acoustic systems that let autonomous machines make sense of environments far messier than any benchmark expects.
What we're building
focusMost autonomous-driving perception research is designed and benchmarked on orderly, car-dominated traffic. Lynx works on the harder, largely unaddressed case: dense, mixed traffic where rickshaws, pedestrians, and vehicles share the road under negotiated rather than lane-disciplined right-of-way — the kind found throughout South Asian megacities like Dhaka. We combine multiple sensing channels to build a picture of the scene that holds up when any single one doesn't.
- Vision
- Multi-camera stereo vision for depth and structure in scenes with no clean sightlines.
- Positioning & motion
- Precise vehicle positioning and inertial sensing, calibrated for streets where GPS alone falls short.
- Acoustic sensing
- Ambient audio as a sensing channel, catching what cameras miss in clutter and glare.
- Fusion
- Combining every channel into one read of the scene, built for the world's hardest traffic, not its easiest.
Research areas
4 activeRobotics
Mobile and field robotics: navigation, manipulation, and control for machines operating outside controlled lab conditions — on unpaved ground, in traffic, and around people who don't follow the rules a robot expects.
Vision
Computer vision for messy, high-density scenes: detection, tracking, and depth estimation tuned for conditions most vision benchmarks don't cover — crowding, occlusion, glare, and unmarked geometry.
Autonomous systems
Perception and decision-making for autonomous vehicles in mixed, unregulated traffic, where lane markings are a suggestion and right-of-way gets negotiated in real time.
Acoustic sensing
Spectral analysis and acoustic event detection that let machines hear what they can't always see — horns, sirens, footsteps, and the ambient signature of a crowded street.
Led by
peopleMollah Md Saif
Adjunct Lecturer, BRAC University — Principal Investigator, Lynx
Leads Lynx's work across robotics, perception, and sensing, with a focus on systems built for real-world conditions rather than curated benchmarks.
Lynx is currently a team of 8 — researchers and students working across robotics, vision, autonomous systems, and acoustic sensing.
Work with us
contactLynx is looking for students and collaborators interested in robotics, computer vision, autonomous systems, or acoustic sensing. If that's you, reach out.