Data augmentation

Data augmentation is done in depth_calibration/image_processing: You can edit this params:

• ##### Augmented data params
• augmented_data_copies = 4 # Number of copies of augmented data that will be created and saved, 0 if you don't want it
• weight_mean = 1.
• weight_dev = 0.05
• biass_mean = 0.
• biass_dev = 9.

Then to each channel of the image random Gaussian noise will be introduced like this:

for each channel:

weight = random.normal(weight_mean, weight_dev)

biass = random.normal(biass_mean, biass_dev)

img[channel] = img[channel]*weight + biass

Note: This function is called introduce_noise and can be found at depth_calibration/depth_helper.py