Adamdad / keras-YOLOv3-mobilenet

I transfer the backend of yolov3 into Mobilenetv1,VGG16,ResNet101 and ResNeXt101

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

About the input_shape of the model

zhangyufei1995 opened this issue · comments

Will the input of the model mobilenet-yolov3 be 320320? But why did you write 416416 here?
image

I am just using a typical input size(like 416 416) to estimate the feature size when write the code. The only constraint of the inputsize is that it must be the multiple of 32

I am just using a typical input size(like 416 416) to estimate the feature size when write the code. The only constraint of the inputsize is that it must be the multiple of 32

I have trained a model with input_size=320, but the test result of input_size=320 is bad than input_size=416, why?

It no surprise that smaller input size always decreases the performance of DL algorithm.

It no surprise that smaller input size always decreases the performance of DL algorithm.
I see, Thank you~