All projects
Computer vision
Pedestrian detection for live video
A detection pipeline taken from photo cameras to video, by reworking how pedestrians are trained and classified.
Results
- faster, from 500 ms to 50 ms a frame
- 10×
The challenge
A computational imaging company couldn't take its pedestrian detection from photo cameras to video: it was far too slow for live frames.
What we built
- Real-time training and classification built on Aggregate Channel Features
- A detection pipeline tuned for video frame rates
Built with
- C++
- Matlab
- Image processing
Have something similar in mind?
Tell us what you're working on and we'll come back with a clear next step.