Collected sources and patterns will appear here. Add from search, explore, or the patterns library.
Image -> List<BoundingBox>
Divide a high-resolution image into overlapping slices, execute object detection on each slice individually, and merge the resulting bounding boxes using Non-Maximum Suppression (NMS) back into global coordinates.
Problem it solves
Detecting extremely small objects in high-resolution images is difficult because downsampling destroys fine-grained spatial features.
Consumes
Emits
The real projects this mechanism was found in. Attribution is the point — this is how the best teams actually do it.