OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
AI Models

Mobile Semantic Segmentation

Mobile Semantic Segmentation is an open-source Python and PyTorch project that trains a MobileNetV2-based U-Net to identify hair regions in face images and includes a script to convert trained models to Core ML for iOS.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Mobile Semantic Segmentation — OpenViST project profile in AI Models.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use it when you need model code and a documented training path for hair masks, especially when the target integration is an iOS app using Core ML.

Who it is for

  • Mobile developers adding hair-region segmentation to face-image applications.
  • Machine-learning practitioners training segmentation models with labeled face images and masks.

Documented uses

  • Train a hair-segmentation model using LFW face images and masks.
  • Convert a trained PyTorch model into Core ML for use in an iOS app.
  • Use a predicted hair mask as a basis for a hair-color preview.
Before you choose

Requirements & limits

The repository is public and specifies the MIT license.

  • The example model uses a MobileNetV2-inspired U-Net architecture.
  • Training instructions specify Python 3.8, LFW images and masks, and a provided training configuration.
  • The repository reports 0.89 IoU for its 224-pixel pretrained model.
  • The Core ML conversion path is documented; TensorFlow Lite conversion for Android is listed as unfinished.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Train a hair-segmentation model

Train the provided model using LFW face images and corresponding masks.

Before you start

  • Python 3.8
  • Dependencies installed with the documented pip command and PyTorch wheel index
  • LFW face images and their masks arranged in the documented folder structure

Steps

  1. Place face images under data/lfw/raw/images and matching masks under data/lfw/raw/masks.

  2. From the src directory, run python run_train.py params/002.yaml.

Expected outcome

The project training script is configured to train the model; for 224-pixel input, pretrained MobileNetV2 weights are automatically downloaded.

Guide limitations

  • The README refers users to issue #11 for obtaining mask images.
  • The reported 0.89 IoU applies to the listed pretrained 224-pixel model and is not a result guarantee for a newly trained model.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
7 November 2026
Original website title
akirasosa/mobile-semantic-segmentation · Official homepage title · retained in its source language
How we review sourcesSuggest a correction