FLOPs and Params - Githubissues

yangxue0827 / RotationDetection

This is a tensorflow-based rotation detection benchmark, also called AlphaRotate.

https://rotationdetection.readthedocs.io/

Apache License 2.0

1.07k stars 180 forks source link

FLOPs and Params #74

Closed mikilerGit closed 2 years ago

mikilerGit commented 2 years ago

Thoese numer is too small .

yangxue0827 commented 2 years ago

Params is correct, whereas Flops are problematic because the input image does not specify a specific size. Therefore, to get the correct Flops, you need to explicitly specify the size of the input image.

mikilerGit commented 2 years ago

ok,I will try to specify the size of the input image. Thanks for your reply.

mikilerGit commented 2 years ago

Hi , How can i specify the size of the input image ? I can not sovle it. Can you give me some advise ?

yangxue0827 commented 2 years ago

example： Img = tf.reshape(Img, [800, 800, 3])

mikilerGit commented 2 years ago

I have got the result 1640347643(1) img shape=（800，800） Is it right ?

yangxue0827 commented 2 years ago

It looks correct