Kaia.ai platform's JS client library
- Sample app
- Sample app source code
- Sample app source code, built with node.js and webpack
let tfMobile = await createTfMobile(model); // load model
...
let result = await tfMobile.run([img], // classify image
{feed: [
{width: size,
height: size,
inputName: 'input',
imageMean: 128.0,
imageStd: 128.0,
feedType: 'colorBitmapAsFloat'
}],
run: {enableStats: false},
fetch: {outputNames: ['MobilenetV1/Predictions/Softmax'], outputTypes: ['float']}
});
let probabilities = result.output[0];
...
tfMobile.close(); // optional
- Sample app
- Sample app source code
- Sample app source code, built with node.js and webpack
let tfLite = await createTfLite(model); // load model
...
let result = await tfLite.run([img], // classify image
{input: [
{width: size,
height: size,
channels: 4,
batchSize: 1,
imageMean: 128.0,
imageStd: 128.0,
type: 'colorBitmapAsFloat'
}],
output: [
{type: 'float',
size: [1, 1001],
}]
});
let probabilities = result.output[0][0];
...
tfLite.close(); // optional
npm install kaia.js
Now you can require/import kaia.js
:
import { TfMobile, TfLite } from 'kaia.js';
dist/kaia.mjs
is a valid JS module.dist/kaia-iife.js
can be used in browsers that don't support modules.idbKeyval
is created as a global.dist/kaia-iife.min.js
As above, but minified.dist/kaia-iife-compat.min.js
As above, but works in older browsers such as IE 10.dist/kaia-amd.js
is an AMD module.dist/kaia-amd.min.js
As above, but minified.
These built versions are also available on jsDelivr, e.g.:
<script src="https://cdn.jsdelivr.net/npm/kaia.js/dist/kaia-iife.min.js"></script>
<!-- Or in modern browsers: -->
<script type="module">
import { createTfMobile, createTfLite } from 'https://cdn.jsdelivr.net/npm/kaia.js';
</script>
and unpkg
<script src="https://unpkg.com/kaia.js/dist/kaia-iife.min.js"></script>
<!-- Or in modern browsers: -->
<script type="module">
import { createTfMobile, createTfLite } from 'https://unpkg.com/kaia.js';
</script>
- To make a custom TfMobile model please follow a detailed Google Codelabs TFMobile tutorial
- To make a custom TfLite model please follow a detailed Google Codelabs TFLite tutorial