MaskRCNN_iOS

An implementation of MaskRCNN based on CoreML.

Introduction

This project implements a MaskRCNN model for person/cat/dog/table/face detection and segmentation which is trained by ourselves.

Also it includes another model for semantic segmentation and depth estimation.

The content is like this:

Performance

Models

To use this project, you need to download the pre-trained models and drag these models into this project in xcode. Or you can put them in ./DPFruit directly.

Also do not forget the opencv2.framework.

Workflow

Take the MaskRCNN as an example, some layers are not supported by CoreML. We implement those parts in c++ and then convert variables between c++ structures(blue parts) and MLMultiArray(yellow parts).

So you can see this:

Contact