The popular way to find objects inside of images is using the mask region convolutional neural network. It’s a fancy way to say “pick a portion of an image and look for something”.
With neural nets like Mask-CNN, it’s possible to get very sophisticated with what you can detect.
While we are not yet at the stage of being able to identify anything, we can do things like animal discernment (which animal is this?).
Here’s a tutorial with reference images that will show you how to pick out the images with kangaroos. It’s based on the matterhorn implementation of R-CNN, so be sure to git clone it before you start the lesson.
The overview for the tutorial is as follows:
The Mask Region-based Convolutional Neural Network, or Mask R-CNN, model is one of the state-of-the-art approaches for object recognition tasks. The Matterport Mask R-CNN project provides a library that allows you to develop and train Mask R-CNN Keras models for your own object detection tasks. Using the library can be tricky for beginners and requires the careful preparation of the dataset, although it allows fast training via transfer learning with top performing models trained on challenging object detection tasks, such as MS COCO.
In this tutorial, you will discover how to develop a Mask R-CNN model for kangaroo object detection in photographs.
After completing this tutorial, you will know:
How to prepare an object detection dataset ready for modeling with an R-CNN.
How to use transfer learning to train an object detection model on a new dataset.
How to evaluate a fit Mask R-CNN model on a test dataset and make predictions on new photos.
Let’s get started.
This tutorial is divided into five parts; they are:
* How to Install Mask R-CNN for Keras
* How to Prepare a Dataset for Object Detection
* How to a Train Mask R-CNN Model for Kangaroo Detection
* How to Evaluate a Mask R-CNN Model
* How to Detect Kangaroos in New Photos
To proceed to the kangaroo detection with Mask R-CNN, go here.
- xml.etree.ElementTree API
- matplotlib.patches.Rectangle API
- matplotlib.pyplot.subplot API
- matplotlib.pyplot.imshow API
- Splash of Color: Instance Segmentation with Mask R-CNN and TensorFlow, 2018.
- Mask R-CNN – Inspect Ballon Trained Model, Notebook.
- Mask R-CNN – Train on Shapes Dataset, Notebook.
- mAP (mean Average Precision) for Object Detection, 2018.