matterport/mask_rcnn

★ 25,559PythonAudience · developerComplexity · 4/5StaleLicenseSetup · moderate

Mindmap

mindmap
  root((repo))
    What it does
      Detect objects
      Draw boxes
      Pixel masks
      Instance segmentation
    Tech stack
      Python
      TensorFlow
      Keras
      ResNet101
    How it works
      Feature Pyramid Network
      Multi-scale scanning
      Pre-trained weights
    Use cases
      Medical imaging
      Robotics
      Autonomous vehicles
      Industrial inspection
    Getting started
      Jupyter notebooks
      Custom training
      COCO dataset

mindmap root((repo)) What it does Detect objects Draw boxes Pixel masks Instance segmentation Tech stack Python TensorFlow Keras ResNet101 How it works Feature Pyramid Network Multi-scale scanning Pre-trained weights Use cases Medical imaging Robotics Autonomous vehicles Industrial inspection Getting started Jupyter notebooks Custom training COCO dataset

Things people build with this

USE CASE 1

Detect and outline objects in medical scans to assist diagnosis and treatment planning.

USE CASE 2

Enable robots to identify and grasp individual items in cluttered environments.

USE CASE 3

Help autonomous vehicles recognize pedestrians, vehicles, and road signs with precise boundaries.

USE CASE 4

Inspect manufactured parts to spot defects and measure dimensions automatically.

Tech stack

PythonTensorFlowKerasResNet101Feature Pyramid Network

Getting it running

Difficulty · moderate Time to first run · 30min

Requires TensorFlow/Keras installation and pre-trained model weights download; GPU optional but recommended for inference speed.

Use freely for any purpose, including commercial use, as long as you keep the copyright notice.

In plain English

Mask R-CNN is a Python implementation of a computer vision technique that can look at a photo and identify every distinct object in it, draw a box around each one, and paint a precise pixel-level outline (called a segmentation mask) around its exact shape. For example, given a street scene it can simultaneously find the cars, people, and traffic lights, label each one separately, and trace their exact outlines rather than just boxing them. This is called instance segmentation, meaning each individual object gets its own mask even if two objects of the same type overlap. The model is built on Keras and TensorFlow, two popular Python frameworks for building and training AI models. It uses a neural network architecture called Feature Pyramid Network combined with a ResNet101 backbone, these are layered mathematical structures that scan an image at multiple scales to catch both large and small objects. The repository includes pre-trained weights from the MS COCO dataset (a large collection of labeled everyday photos), Jupyter notebooks for visualization, and tools to train the model on your own custom dataset. Researchers and developers would use this when they need to detect and precisely outline objects in images or video, such as in medical imaging, robotics, autonomous vehicles, or industrial inspection. It requires Python 3, Keras, and TensorFlow.

Copy-paste prompts

Prompt 1

Show me how to use Mask R-CNN to detect and segment objects in my own images using the pre-trained COCO weights.

Prompt 2

How do I fine-tune Mask R-CNN on a custom dataset of images for my specific use case?

Prompt 3

Walk me through the Jupyter notebooks in this repo to understand how instance segmentation works.

Prompt 4

What's the difference between Mask R-CNN and regular object detection, and when would I use each one?

Prompt 5

Help me set up Mask R-CNN with TensorFlow and Keras to run inference on video frames.

Open on GitHub → Explain another repo

Generated 2026-05-18 · Model: sonnet-4-6 · Verify against the repo before relying on details.