Get started with convnetjs CDN
MIT licensed
ConvNetJS: Lightweight web library for real-time image classification via convolutional neural networks.
Tags:- machine
- learning
- AI
- convnet
- neural
- network
- networks
- convolutional
- deep
Stable version
Copied!
How to start using convnetjs CDN
<!DOCTYPE html>
<html>
<head>
<title>Get started with convnetjs CDN - cdnhub.io</title>
<script src="https://cdn.cdnhub.io/convnetjs/0.3.0/convnet-min.js"></script>
</head>
<body>
<script>
// Define the input shape and number of output classes
const inputShape = [28, 28, 1];
const numClasses = 10;
// Load the MNIST dataset
const net = new NeuralNetwork();
net.addInput('input', inputShape);
net.addLayer('conv1', new ConvLayer({
input: 'input',
output: 32,
kernelSize: 3,
padding: 'same',
activation: 'relu'
}));
net.addLayer('pool1', new MaxPoolingLayer({ input: 'conv1.output', poolSize: [2, 2] }));
net.addLayer('conv2', new ConvLayer({
input: 'pool1.output',
output: 64,
kernelSize: 3,
padding: 'same',
activation: 'relu'
}));
net.addLayer('pool2', new MaxPoolingLayer({ input: 'conv2.output', poolSize: [2, 2] }));
net.addLayer('flatten', new FlattenLayer({ input: 'pool2.output' }));
net.addLayer('fc1', new FullyConnectedLayer({
input: 'flatten.output',
output: 128,
activation: 'relu'
}));
net.addLayer('output', new SoftmaxLayer({ input: 'fc1.output', numClasses }));
net.addOutput('output');
net.loadWeights('https://cdn.cdnhub.io/convnetjs/examples/mnist/mnist_model_data.json');
// Preprocess the input image and predict the output class
const inputImage = new Tensor(inputShape, 'float32');
inputImage.set(0.1); // Set all pixels to 0.1 for simplicity
net.predict(inputImage).then(output => {
const predictedClass = output.argMax();
console.log(`Predicted class: ${predictedClass}`);
});
</script>
</body>
</html>
Copied!
Copied!