search

BichenWuUCB / squeezeDet

A tensorflow implementation for SqueezeDet, a convolutional neural network for object detection.
BSD 2-Clause "Simplified" License
739 stars 306 forks source link

Data augmentation oddities #98

Closed munum closed 6 years ago

munum commented 6 years ago

Hi I was poking around the code for image augmentation and this part caught my eyes. max_drift_x and max_drift_y seem to address the issue of pushing the bbox to the -x and -y territories. But how about the other directions? Shouldn't dx be in the form of

dx = np.random.randint(max(-mc.DRIFT_X, -VALUE), min(mc.DRIFT_X+1, max_drift_x))