cherrypiecoco / maskrcnn-extension

An extension of Mask RCNN model for training on custom data or masking videos rather than images

Step 1 - Create a Conda virtual environment

We will be using Anaconda with python 3.6. If you don't have Anaconda, follow this tutorial https://www.youtube.com/watch?v=T8wK5loXkXg

Create a anaconda virtual environment by:

conda create -n MaskRCNN anaconda python=3.6

Enable environment support in jupyter:

conda install nb_conda

Step 2 - Install the Dependencies

Activate the virtual environment, install dependencies and run setup:

conda activate MaskRCNN
conda install jupyter
pip3 install -r requirements.txt
python3 setup.py install

Step 3 - Install pycocotools

Needed to use Microsoft CoCo dataset:

pip3 install git+https://github.com/waleedka/coco.git#subdirectory=PythonAPI

Step 4 - Launch Jupyter

jupyter notebook

Step 5 - Open MaskRCNN Notebook

Everything is in MaskRCNN.ipynb.

Note: Video processing is in Video_Masker.ipynb.