Distilling Image Classifiers in Object Detectors - Supplementary Material - S1 Code for Our Approach

Neural Information Processing Systems 

The classification-related errors typically drop by using our classification distillation strategy. See, for example, the classification error for IoUs smaller than 0.65, and the error of both cls and loc for all IoUs.

Similar Docs  Excel Report  more

TitleSimilaritySource
None found