data augmentation for caffe in mobilenet SSD

[siri]

Yolo configuration provides you all these parameters while training so the model itself add augmentation you images & train the model:

1. Addition of Gaussian noise.
2. Brightness variation.
3. Addition of Salt and Pepper noise.
4. Scaling of the image.
5. Affine rotation given the maximum possible rotation.
6. Perspective Transform within user provided min and max rotation angle about x, y and z axes.
7. Image sharpening.
8. Power Law transfrom for illumination effects. (needs to be update)

How do I do this thing for MobileNet SSD in caffe? Or I need do it separately manually & then pass all those images for training. I'm using this: https://github.com/chuanqi305/MobileNet-SSD