training of mask rcnn on self created dataset using python and tensorflow

Basic system reqirement

1. windows os
2. nvidia gpu
3. installation of CUDA
4. installation of CUDAnn
5. Anaconda
6. python 3

step 1 clone the above repository and extract it into your local system and also extract the dataset folder and place it into project directory and open cmd in that directory

step 2 download the pre trained coco weight from the given link below. https://github.com/matterport/Mask_RCNN/releases and select mask_rcnn_coco.h5 from mask rcnn 2.0 version and place it in the project directory.

step 3 creation of virtual conda enviornment open cmd and run the command conda create -n myenv

step 4 enter into the virtual enviornment conda activate myenv

step 5 install all the dependencies by using following command pip install tensorflow-gpu==1.13.1

pip install q keras==2.1.0

pip install -r requirements.txt

step 6 Run setup from the repository root directory

python setup.py install

• if you donot have jupyter notebook install in your enviornment use this command pip install notebook

step 6 open jupyter notebook using command jupyter notebook

and navigate to the mines.ipynb file and press shift + enter to execute that a code block.