AugmentationTutorial
principle
gray scale and rgb scale compatability
detection(with boxes) and classification(with label) compatability
cpp and python framework
progress
- flip (vertical and horizontal done)
- rotate 90 degree
- distort/color jitter
- expand
- random crop
- clip bbox to border
- resize
- stable resize (keep width/height ratio unchanged)
- data validation check
- data format convert, Tensor to numpy in pyTorch, tf.Variable to numpy in Tensorflow, etc.
- check data batch and display
- Basic format for test
- cpp data augmentation framework