SightHub

SightHub was a startup I co-founded with the goal of giving brick-and-mortar shops the kind of visitor analytics that online shops take for granted. Using a Raspberry Pi and a camera placed at the entrance, the system live-recorded foot traffic and periodically sent frames to the SightHub server for processing. Shop owners could then see how many people were in their store at any given time, and build up insights into when traffic peaks and troughs occur throughout the day.

We built an executable pipeline where a frame could enter the system and pass through multiple stages — resizing, greyscale conversion, face detection, face recognition, and face tracking — each stage doing a specific operation before handing off to the next. The processed output was synced back to the API as structured data. Building something that could take raw video and turn it into real business intelligence was genuinely exciting.

This project later inspired my graduation thesis: Multi-Frame Face Reconstruction With a 3D Morphable Model. Ask me about it if you're curious.

SightHub pushed me deep into computer vision for the first time. Working with Python, Django, OpenCV, NumPy, and FFMpeg on a real product gave me a solid foundation in image processing pipelines. I also learned about HOG — Histograms of Oriented Gradients — and how classical computer vision techniques compare to the deep learning approaches that would follow. Running a startup alongside my studies was a formative experience in moving fast under constraints.