|Assignee:||S. Wrede||% Done:|
|Category:||-||Estimated time:||4.00 hours|
Implement a component which extract regions-of-interest patches (based on bounding boxes) out of a video stream, e.g., to generate training data for face or object recognizers. Initial implementation should at least work with rst.vision.Faces.
- should utilize the timesync library to fuse bounding box detections with the closest video image based on create timestamps (or using cause vector information)?
- should write the patches to files.