Practical overview
When it fits
Use the repository to explore MobileNet image predictions in Core ML, compare its fixed-image and live-camera demos, or examine its documented model-conversion process.
Who it is for
- iOS developers exploring image predictions with MobileNet and Core ML.
- Developers looking for a reference project with both a still-image demo and a live-camera demo.
- Readers interested in the documented Caffe-to-Core-ML model conversion process.
Documented uses
- Inspect a Swift implementation of MobileNet for Apple's Core ML framework.
- Run a demonstration that predicts from a cat picture.
- Run a device-based demonstration that predicts from live camera video.
- Follow the repository's documented conversion procedure to create MobileNet.mlmodel from a Caffe model.
Requirements & limits
- The repository includes the pretrained MobileNet.mlmodel, so model conversion is not required to use the provided model.
- The Cat Demo is documented for Xcode 9 and iOS 11 or the simulator.
- The Camera Demo requires a device and is documented as incompatible with the simulator.
- The conversion instructions require Python 2.7 and Keras 1.2.2.
Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.
Run the included image-prediction demos
View the Cat Demo prediction or run the live-camera prediction demo.
Before you start
- Xcode 9
- For the Cat Demo, a device with iOS 11 or the simulator
- For the Camera Demo, an iOS device; the simulator is not supported for this demo
Steps
Open the project in Xcode 9.
Run the Cat Demo on an iOS 11 device or in the simulator to view its cat-picture prediction.
Run the Camera Demo on a device to see predictions from a live video feed.
Expected outcome
The Cat Demo displays a prediction for a cat picture; the Camera Demo makes predictions from live video as often as it can manage.
Guide limitations
- The Camera Demo does not work in the simulator.
- The instructions specify Xcode 9 and iOS 11 for the Cat Demo.
Convert the Caffe model to Core ML
Recreate the included MobileNet.mlmodel from the source Caffe model.
Before you start
- The MobileNet-Caffe caffemodel file downloaded into the project’s top-level folder
- The repository's bundled mobilenet_deploy.prototxt
- A virtual environment created with /usr/bin/python2.7
- TensorFlow, Keras 1.2.2 and coremltools
Steps
Download the caffemodel file into the project’s top-level folder; the repository says the deployment prototxt is already included.
Create and activate a virtual environment using /usr/bin/python2.7.
Install TensorFlow, Keras 1.2.2 and coremltools in that environment.
Run python coreml.py, then deactivate the virtual environment.
Expected outcome
The documented script creates MobileNet.mlmodel.
Guide limitations
- The instructions warn that using a Python version other than /usr/bin/python2.7 causes the conversion script to crash.
- The instructions require Keras 1.2.2 rather than Keras 2.0.
- The provided MobileNet.mlmodel means conversion is unnecessary if the goal is simply to use the included model.
Sources & review
Source record details
- Source availability
- Active
- Sources checked
- 7 November 2026
- Original website title
- hollance/MobileNet-CoreML · Official homepage title · retained in its source language
Related options
Forge — The repository identifies Forge as a separate neural-network library for iOS 10 that includes a Metal implementation of MobileNet.
Explore project →