-
First approach is quickly iterating from scratch using Pytorch and resnet26d. SCORE-> 0.7
-
Second approach is using Fastai API to make the process quicker and resnet50: SCORE-> 0.18 (Definitely did something wrong predicting, since the error_rate was about 0.11).
-
Third approach is using a more complex model in the first notebook. SCORE-> 0.8
-
Fourth approach is using Fastai API to make the process quicker and implement Callbacks, easier Progressive Resizing, and Test-Time-Augmentation.
SuperDXCEL/PaddyKaggle
Folders and files
| Name | Name | Last commit date | ||
|---|---|---|---|---|