ML Classifier is a React front end for a machine learning engine for quickly training image classification models in your browser. Models can be saved with a single command, and the resulting models reused to make image classification predictions.
This package is the UI front end for ml-classifier.
A walkthrough of the code can be found in the article Image Classification in the Browser with Javascript.
An interactive demo can be found here.
Screenshot of demo
ml-classifier-ui can be installed via yarn or npm:
yarn add ml-classifier-ui
or
npm install ml-classifier-ui
You can fork a live running version at codesandbox.io.
Start by instantiating a new MLClassifierUI.
import React from 'react';
import ReactDOM from 'react-dom';
import MLClassifierUI from 'ml-classifier-ui';
ReactDOM.render(<MLClassifierUI />, document.getElementById('root'));
MLClassifierUI accepts a number of parameters:
- getMLClassifier (
Function) Optional - A callback that returns an instance of the underlyingml-classifierobject. Call this if you want to programmatically call methods likeaddData,train, andpredict. For more information onml-classifier's API methods refer to it's documentation. - methodParams (
Object) Optional - A set of parameters that will be passed in calls toml-classifier's methods. See below for more information. - uploadFormat (
string) Optional - A string denoting what type of upload format to accept. Formats can beflatornested. See below note for more information on that. If omitted, all formats are accepted. - imageFormats (
string[]) Optional - An array of file extensions to accept. By default, all valid images are accepted. Images are transformed via the nativeImagetag in the browser, so if the browser can display the image, it'll be processed. - showDownload (
boolean) Optional - A flag denoting whether to show a download button or not. Defaults to true.
MLClassifierUI also accepts a number of callbacks that are called on the beginnings and ends of ml-classifier functions. You can view a list of those here.
getMLClassifier returns an instance of ml-classifier for programmatic access to the underlying methods.
<MLClassifierUI
getMLClassifier={(mlClassifier) => {
mlClassifier.addData(...);
}}
/>
methodParams can be used to pass method-specific parameters to ml-classifier. The key will be used to determine which method to pass parameters to.
Accepted keys are train, evaluate, and save. Other keys will be ignored.
<MLClassifierUI
methodParams={{
train: {
epochs: 20,
},
evaluate: {
batchSize: 32,
},
save: {
},
}}
/>
uploadFormat corresponds to how uploaded images should be organized. There are two options:
Expects images to be organized in folders matching the label. Only the immediate parent folder's name will be used as the label. For example:
- containing-folder/
- dogs/
- IMG-1.jpg
- IMG-2.jpg
- IMG-3.jpg
- cats/
- IMG-1.jpg
- IMG-2.jpg
- IMG-3.jpg
Will product an array of three dogs labels and three cats labels.
Nested folders will be searched recursively, but only immediate parent folders' names will be used. If an invalidly nested structure is found an error will be thrown.
Expects files' names to be the label. Nested folders will be searched recursively (if the browser supports it) to build a flat array of files.
- folder/
- dog-1.jpg
- dog-2.jpg
- dog-3.jpg
- cat-1.jpg
- cat-2.jpg
- cat-3.jpg
<MLClassifierUI
uploadFormat={"nested"}
/>
imageFormats denotes the list of acceptable image formats for upload. Any images not matching the list of acceptable formats will be ignored.
<MLClassifierUI
imageFormats={[
'png',
'gif',
]}
/>
Contributions are welcome!
You can run the local example with:
yarn watch
ml-classifier-ui is written in Typescript and React.
Tests are a work in progress. Currently, the test suite only consists of unit tests. Pull requests for additional tests are welcome!
Run tests with:
yarn test
This project is licensed under the MIT License - see the LICENSE file for details