The following project was built on top of the FAIR's detectron2. The primary objective was to detect and segment very small objects in the image. To do that, we made some changes at the algorithm level and some at the data level.
At data level, we added a utility to slice the training images into 3x3 grid. This helps us retain the small objects in the image. Moreover, we get an enhanced region of interest.
At algorithm level, using the Faster RCNN, we made several modifications:
- Added a smaller anchor box (16px) and attached it to the early stage of the backbone network (ResNet-50).
- Increased input resolution from 800 to 1280. We need to keep details in order to detect small objects.
- Added SAHI at the inference.
iMaterialist (Fashion) 2020 at FGVC7
Fine-grained segmentation task for fashion and apparel
ResNet-50
Original Image
Low resolution (600px) | High resolution (960px) |
---|---|
![]() |
![]() |
Without small anchor size | With small (16px) anchor size |
---|---|
![]() |
![]() |