Home

Introduction

mmcv is a collection of common methods and utils for computer vision. It consists of the following modules.

io: It provides a set of unified APIs for loading and dumping data, supporting different file formats. It also provides some wrappers for common file and directory operations.
image: It provides friendly APIs for image IO and processing, based on OpenCV (compatible with multiple versions).
video: It provides sequence-like interfaces to manipulate videos, and some video processing tools.
visualization: It provides some visualization methods to display images and various annotations.
utils: It is a collection of useful utils, including timer, progress bar, etc.
config: A helper class for the usage of config files, supporting py/yaml/json.
torchpack: It defines a flexible training pipeline controlled by built-in and custom hooks, and also provides some useful methods.
backbones: Implementation of popular backbones.