Vision Detector

Vision Detector app execute image processing with CoreML model on iPhone/iPad.
Usually, CoreML models must be previewed on Xcode or an application must be built with Xcode to run on the iPhone.
With Vision Detector, you can easily run CoreML models on your iPhone.

Using CreateML or coreml tools, prepare a machine learning model in CoreML format that you wish to run.
Copy the machine learning model into the iPhone/iPad file system. The file system is the area visible from the iPhone's "File" application, either in the local device or in various cloud services (iCloud Drive, One Drive, Google Drive, DropBox, etc.). You can also use AirDrop, etc.
Launch the app, select and load the machine learning model.
Select the input source image from;
- Video from iPhone/iPad built-in camera
- Still image from the built-in camera
- Photo library
- File system
In the case of video, a continuous inference is performed on the camera image, but the frame rate and other parameters depend on the performance of the device.
Supported machine learning models are;
- Image classification
- Object detection
- Style transfer.
Models that do not have a non-maximum suppression layer or models that input/output data in the form of a MultiArray are not supported.

In the local documents folder of the iPhone, there will be a folder entitled Vision Detector, which will contain an empty CSV file named customMessage.csv.
In this file, you can define a custom message to be displayed in the object detection video processing.
(Label output by YOLO, etc.),(Message)
(Label output by YOLO, etc.),(Message)
and describe the table data with 2 columns as shown above.

This application does not include a machine learning model.