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.
- Pytorch version 0.4 or higher.
Given a test image, the trained model outputs the instance masks in the image:
- Download the checkpoint from here and add it to folder
checkpoints
:
https://drive.google.com/open?id=19aZJ3MQxZ3sdXlwAy4yK-TzFr4l88o7b
- Evaluate the trained mask rcnn on the PASCAL validation set,
python test.py
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 |
---|---|
Run a mask rcnn on PASCAL 2012 with the following command:
python train.py
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.
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}
}