Explore more single-scene augmented training approaches
Chintan2108 opened this issue · comments
vesper2108 commented
- Currently the network is being trained on a single scene which has been rotated at 0, 90, -90 and 180 degrees.
- Explore more augmentation strategies such as contrast augmentation, skew augmentation, flip augmentation
- Try and train with various augmented data and record the best results from amongst all