Skip to content

Latest commit

 

History

History
17 lines (10 loc) · 1.06 KB

README.md

File metadata and controls

17 lines (10 loc) · 1.06 KB

Focal-Loss-based-classification

Image Classification

Image Classification is performed on CIFAR-10 dataset with both Categorical Cross Entropy Loss(CCE) and Focal Loss(FL). Focal Loss showed better performance when compared to CCE. The training loss is much less for FL in each epoch compared to CCE. This is primarly because FL imposes heavy loss when correct samples are classified wrong.

Focal Loss is perfomed with two sets of values:

alpha = 0.25,gamma=2.

alt text

alt text

alpha = 1,gamma=0.

alt text

alt text