Grape-Segmentation

A Machine Learning course final project.In this project, we aim to use instance segmentation to detect the grapes in the vineyards.
We augmented the Embrapa Wine Grape Instance Segmentation Dataset and applyed Mask R-CNN on the augmented dataset to generalize the trained model to get resultful detection on the full-plant grape photos from Vitis International Variety Catalogue.

Running Environment

Make sure:

• Tensorflow==1.15.2
• Keras==2.1.6
• Clone the wgisd dataset to the current folder by git clone https://github.com/thsant/wgisd.git
• The code of mrcnn module is already in the directory

Quick Start

Use Trained Model to Detect

Open detection.ipynb and run all cells.
Or, if you have your own grape full plant images, you can change IMAGE_DIR to your directory and run the code.

Train the Mask R-CNN on Dataset

Data Augmentation

First, make sure that data_augmentation.ipynb is inside wgisd folder.
Then open data_augmentation.ipynb and run all cells.

Train and evaluate the model

Open grapes.ipynb and run all cells.