Awesome

ServiceNow completed its acquisition of Element AI on January 8, 2021. All references to Element AI in the materials that are part of this project should refer to ServiceNow.

WISE - BMVC 2019

Where are the Masks: Instance Segmentation with Image-level Supervision

[Paper]

Requirements

• Pytorch version 0.4 or higher.

Description

Given a test image, the trained model outputs the instance masks in the image:

Checkpoint for the weakly supervised mask rcnn

1. Download the checkpoint from here and add it to folder checkpoints:

https://drive.google.com/open?id=19aZJ3MQxZ3sdXlwAy4yK-TzFr4l88o7b

2. Evaluate the trained mask rcnn on the PASCAL validation set,

python test.py

Test on a single image

Run a trained mask rcnn on a single image as follows:

python test_on_image.py

The expected output is shown below, and the output image will be saved in the same directory as the test image.

ground-truth	predictions

Training

Run a mask rcnn on PASCAL 2012 with the following command:

python train.py

Class-agnostic proposals

Proposals can be obtained from

• deepmask: https://github.com/facebookresearch/deepmask
• MCG:https://github.com/jponttuset/mcg
• COB: http://www.vision.ee.ethz.ch/~cvlsegmentation/cob/code.html

They all have the same supervision which is class-agnostic mask labels from a possibly different training set. For unsupervised proposal-based method, use selective search.

Citation

If you find the code useful for your research, please cite:

@article{laradji2019masks,
  title={Where are the Masks: Instance Segmentation with Image-level Supervision},
  author={Laradji, Issam H and Vazquez, David and Schmidt, Mark},
  journal={arXiv preprint arXiv:1907.01430},
  year={2019}
}