image-labelling-tool
image-labelling-tool copied to clipboard
Vertical flip?
In train_utils.py, you've made a horizontal flip:
` # 5. Random horizontal flipping
img, params = transforms.random_flip(
img, x_random=True, return_param=True)
bbox = transforms.flip_bbox(
bbox, (self.size, self.size), x_flip=params['x_flip'])
`
I'm wondering if there is a reason why you didn't also implement a vertical flip (with a #6 paragraph, copy of #5 with x_random
replaced by y_random
, and x_flip
by y_flip
), which would again double the chances of recognizing objects.