MindSpore Vision Documents

MindSpore Vision is an open source computer vision research toolboxbased on MindSpore framework, which includes classification, detection (to do), video (to do) and 3D (to do). MindSpore Vision is designed to serve the growing research community by providing a flexible toolkit to reimplement existing methods and develop their own new methods. We also hope to invite you to participate in our development of MindSpore Vision.

Warning

MindSpore Vision is about to be abandoned, and mindvision in Pypi only supports MindSpore >=1.7.0 <=1.8.1. If you want to reimplement existing methods and develop your own new methods or to participate in our development of Computer Vision research of MindSpore, please use MindCV.

Scenarios where MindSpore Vision is applied

• Classification

  Image classification toolbox and benchmark.

Base Structure

MindSpore Vision is a MindSpore-based Python package that provides high-level features:

• Classification: Deep neural networks work flows built on an engine system.

• Backbone: Base backbone of models like ResNet and MobileNet series.

• Engine: Callback functions for model training.

• Dataset: Domain oriented rich dataset interface.

• Utils: Rich visualization and IO(Input/Output) interfaces.

Installation

• Install MindSpore Vision

API References

• mindvision.classification
• mindvision.engine
• mindvision.utils