Randomize_input_shape_period in augmentation config

Accelerated Computing Intelligent Video Analytics TAO Toolkit
1

Hi,
I have a question about “randomize_input_shape_period” parameter mentioned in YOLOV3 documentation:
https://docs.nvidia.com/tao/tao-toolkit/text/object_detection/yolo_v3.html#augmentation-config

As far as I know, all input images should have the same size. But according to this parameter, the height and width of the input images change randomly per K batches.
Is the same size of the input images violated?

2

For this parameter, it is similar to

  • random=1 - randomly resizes network for each 10 iterations from 1/1.4 to 1.4 (data augmentation parameter is used only from the last layer)

https://raw.githubusercontent.com/AlexeyAB/darknet/master/cfg/yolov4.cfg and
CFG Parameters in the different layers · AlexeyAB/darknet Wiki · GitHub

3

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.