This is an HTTP service wrapper for BASNet: Boundary-Aware Salient Object Detection code

The deploy folder contains configuration files for deployment as serverless container with Knative.

It's highly recommended to run this image on a machine with a CUDA compatible Nvidia Card and minimum 6Gb of RAM.

docker run --rm -p 8080:80 docker.io/cyrildiagne/basnet-http

If you're having empty response issues, make sure your docker instance has 6gb of RAM as mentioned here

curl -F "[email protected]" http://localhost:8080 -o result.png

Q: I get an empty response, or "out of memory"

A: Increase your Docker RAM limit to at least 6GB: cyrildiagne/ar-cutpaste#26 (comment)

Q: I get a file that is 256x256 instead of my input image's size

A: That's the right output. You need to resize it to your input image's width and height. See: #11

• Clone this repository: git clone https://github.com/cyrildiagne/BASNet-http.git
• Go into the cloned directory: cd BASNet-http
• Clone the BASNet repository
• Download the pretrained model basnet.pth
• Put the file inside the BASNet/saved_models/basnet_bsi/ folder.

Requires Python v3.6+

virtualenv venv
venv/bin/activate

pip install torch==0.4.1
pip install -r requirements.txt

python main.py

After you've retrieved the BASNet model.

Download Resnet checkpoint

curl https://download.pytorch.org/models/resnet34-333f7ec4.pth -o resnet34-333f7ec4.pth

docker build -t basnet .
docker run --rm -p 8080:80 basnet