FixRes is an image scaling strategy that seeks to optimize classifier performance. It is motivated by the observation that data augmentations induce a significant discrepancy between the size of the objects seen by the classifier at train and test time: in fact, a lower train resolution improves the classification at test time! FixRes is a simple strategy to optimize the classifier performance, that employs different train and test resolutions. The calibrations are: (a) calibrating the object sizes by adjusting the crop size and (b) adjusting statistics before spatial pooling.
Source: Fixing the train-test resolution discrepancyPaper | Code | Results | Date | Stars |
---|
Task | Papers | Share |
---|---|---|
Image Classification | 10 | 31.25% |
Fine-Grained Image Classification | 5 | 15.63% |
General Classification | 4 | 12.50% |
Semantic Segmentation | 2 | 6.25% |
Document Image Classification | 2 | 6.25% |
Self-Supervised Learning | 1 | 3.13% |
Document Layout Analysis | 1 | 3.13% |
Efficient ViTs | 1 | 3.13% |
Classification | 1 | 3.13% |
Component | Type |
|
---|---|---|
🤖 No Components Found | You can add them if they exist; e.g. Mask R-CNN uses RoIAlign |