This is an implementation to matterport's Mask_RCNN repository. Using python3, Keras, and tensorflow, this model detects, segments and builds bounding boxes around objects.
The repository contains:
- All files contained in the parent repository
- An added "CoinCounter" class, which counts the total value of coins in a photo
- A "Blueprint" directory, which contains a blackbox blueprint_class and a blackbox blueprint_inspect_data .py file.
The goal of this repository is to create a system for generalized and easily configurable object detection and segmentation. The blueprint class streamlines the processes of training, testing and inference.
-0. Clone this repository.
-1. Download the COCO weights. You can optionally download the coin and balloon dataset as well. Place the coco weights in your repository and create a directory called "datasets" for the datasets.
-2. Create a "logs" file to save training progress (weights after every epoch and scalers for tensorboard).
-3. Install all of the dependencies > pip3 install -r requirements
-4. Run the setup.py file > python3 setup.py install
