Data Augmentation For Object Detection using Pytorch and PIL (image from VOC dataset)
- Adjust Contrast
- Adjust Brightness
- Adjust saturation
- Lighting Noise
- Flip
- Rotate
- Random crop
- Zoom out (expand image)
- Rotate only bouding box (optional)
- Cutout
- Mixup
I wrote a repo: Implementation of Single Shot MultiBox Detector (SSD) using some data augmentation strategies. Check it out!