Mask-RCNN

Basics and Hand-On NOTE : All the actions and works have been done on Ubuntu Virtual Machine

TO INSTALL Mask_RCNN :

OPEN TERMINAL, TYPE THE FOLLOWING sudo su (enter ur password)

git clone https://github.com/matterport/Mask_RCNN.git (without '#')

cd Mask_RCNN

apt-get install python-pip

apt-get install python3-pip

hash -d pip [Incase of error with pip]

pip install -r requirements.txt

pip3 install imgaug

python setup.py install

pip show mask-rcnn

Now, download the weights (https://github.com/matterport/Mask_RCNN/releases/download/v2.0/mask_rcnn_coco.h5)

Copy the downloaded weights file to the Mask_RCNN/samples/ folder, Now, make a new folder named 'sandbox' in Mask_RCNN folder. This sandbox folder would be the one that would contain all the files that you are going to work with and also the notebooks that would be created.

Working with Custom Datasets and Classes

There are a total of 81 predefined and pretrained classes available in MaskRCNN. To use maskrcnn for classes apart from these 81 classes, follow the steps mentioned in the document : https://docs.google.com/document/d/1wwZPsBMmVs4u0Lf-mHKBjEzvm7FPvcomLftiDb2zIfU/edit?usp=sharing.