This Repository combines the MASK_RCNN algorithm for instance segmentation and DeepLabV2 Algorithm for semantic segmentation in order to produce predictions for the Panoptic Segmentation Challenge.

Alt text

In order to use, it's require to setup the repository first:


Make sure you have Python 3.6 and pip3 9.0.1 and pipenv (other pip versions may work, however they are untested) This also requires CUDA 9 and cuDNN to be installed on system.


  1. git clone to clone the repository

  2. Enter Root directory:

    cd PanopticSegmentation
  3. Install dependencies:

    pipenv install


From inside the root directory:

  1. Enter Virtual Env:

pipenv shell

  1. Run the command:

pipenv run start

  1. Follow the instructions in the menu to run either inference or evaluation of the dataset.