/demo_using_mask_rcnn

This tutorial is one of the demos used in the course “Introduction to Machine Learning” at MINES ParisTech - PSL Research University, lectured by Simon Tamayo. Acknowledgement: This demo was adapted from the Matterport tutorial on Mask_RCNN which implements Mask R-CNN. Mask R-CNN efficiently detects objects in an image while simultaneously generating a high-quality segmentation mask for each instance (developed by K. He, G, Gkioxari, P. Dollár & R. Girshick from Facebook AI Research).

Primary LanguageJupyter Notebook

Image detection and classification using Mask R-CNN

This tutorial is one of the demos used in the course “Introduction to Machine Learning” at MINES ParisTech - PSL Research University, lectured by Simon Tamayo.

This code uses a deep convolutional neural network, and more precisely Mask R-CNN to detect and classify the objects present in an input image. Mask R-CNN does the following:

mask_rcnn_overview *image borrowed from the original paper presenting mask rcnn

Here is what Mask R-CNN does:

  1. From an input_image:

input_image_ex

  1. You can identify the objects in the image, classify them and also obtain their masks:

input_style_ex

Acknowledgement

This demo was adapted from the Matterport tutorial on Mask_RCNN which implements Mask R-CNN. Mask R-CNN efficiently detects objects in an image while simultaneously generating a high-quality segmentation mask for each instance (developed by Kaiming He, Georgia Gkioxari, Piotr Dollár, Ross Girshick).

Getting Started

Before running this code make sure you have Python3 and the packages: numpy, scipy, Pillow, cython, matplotlib, scikit-image, tensorflow>=1.3.0, keras>=2.0.8, opencv-python, h5py, imgaug.

Please note that in order to run this code you also need to download the pre-trained COCO weights (mask_rcnn_coco.h5) from here, and place it in the root directory of this code.

MASK R-CNN

This tutorial uses the deep neural network MASK R-CNN developed by K. He, G. Gkioxar, P. Dollar and R. Girshick from Facebook AI Research (FAIR). Mask R-CNN is an instance segmentation model that can identify pixel by pixel location of any object, returning a classification result and a mask for the location of the instance in the image.

Built With

  • Mask R-CNN - Deep neural network
  • COCO - Image data set used to train Mask R-CNN

This project is licensed under the MIT License - see the LICENSE.md file for details