Home

Awesome

Semantic Amodal Segmentation dataset and API

This is the Python API code for the amodal segmentation dataset proposed in Semantic Amodal Segmentation (CVPR 2017). This API code is built on COCO API.

setup

  1. git clone and compile:
  1. create soft link for coco/bsds images:
  1. dowload annotation files and untar.

notebook demo

  1. To see the annotation and some useful APIs, please run the ipython notebook demo.

evaluate

  1. dowload the baseline amodalMask output on coco val set and untar:

  2. run the segmentation evaluation.

It measures amodal segment proposal quality using average recall. Please see details in table 3a and section 5.1 from the paper.

annotation tool

We also release the web tool we used for annotation in another repo here. It's modified based on OpenSurface.

citation

If you find this dataset useful to your research, please consider citing:

@inproceedings{zhu2017semantic,
    Author = {Zhu, Yan and Tian, Yuandong and Mexatas, Dimitris and Doll{\'a}r, Piotr},
    Title = {Semantic Amodal Segmentation},
    Booktitle = {Conference on Computer Vision and Pattern Recognition ({CVPR})},
    Year = {2017}
}