search

experiencor / keras-yolo2

Easy training on custom dataset. Various backends (MobileNet and SqueezeNet) supported. A YOLO demo to detect raccoon run entirely in brower is accessible at https://git.io/vF7vI (not on Windows).
MIT License
1.73k stars 784 forks source link

imgaug ? #390

Open amapic opened 5 years ago

amapic commented 5 years ago

From where do you get imgaug ? Thank you. I can't find it for anaconda. I need to install it myself ?

robertlugg commented 5 years ago

I believe they are from here. That site does have install instructions. Yes, you would need to get it yourself. I simply removed that code from preprocessing, but you doing that would depend on your specific application.

amapic commented 5 years ago

Thanks a lot !

abhijithvnair94 commented 5 years ago

You can directly install imgaug to your conda enviornment from here. I tried this and it is working finely

amapic commented 5 years ago

ok thanks

dhiren-hamal commented 2 years ago

It was such a pain in my ass to run this project because of imgaug package. Finally, I am able to run this project. Here are my insights. If you want to run this project above tensorflow-gpu=1.3.0 or planning to use >= python3.5then you can't use imgaug based preprocessing, so it's better to remove it.

What I did..

  1. docker pull floydhub/tensorflow:1.3.0-gpu-py2_aws.12
  2. install opencv-python 4.0.21
  3. install imageio 2.6.x
  4. pip install imgaug --no-dependencies
  5. scikit-image 0.14.2

Hope it will be useful.