retrieve json from KNN Classifier instead of saving

[micuat]

→ Description 📝

• have an idea for a new feature 🆕

getting json object instead of saving directly to a file would be helpful when sending data to node.js, for example.

→ Screenshots 🖼

(Any relevant screenshots, sketches, or helpful concept diagrams)

I'm using this makeshift for now, but it should be included in the ml5 library (I think).

const dataset = knnClassifier.knnClassifier.getClassifierDataset();
if (knnClassifier.mapStringToIndex.length > 0) {
  Object.keys(dataset).forEach((key) => {
    if (knnClassifier.mapStringToIndex[key]) {
      dataset[key].label = knnClassifier.mapStringToIndex[key];
    }
  });
}
const tensors = Object.keys(dataset).map((key) => {
  const t = dataset[key];
  if (t) {
    return t.dataSync();
  }
  return null;
});
console.log({dataset, tensors});

→ Helpful Information 🦄

A list of relevant information for this issue. This will help people navigate the project and offer some clues of where to start. Depending on the nature of this issue, important and helpful information might include:

• Web browser and version any
• Operating System any
• ml5 version you're using: 0.1.2

[micuat]

and of course loading

socket.on('lastModel', (data) => {
  // console.log(data)

  if (data) {
    const { dataset, tensors } = data;
    knnClassifier.mapStringToIndex = Object.keys(dataset).map(key => dataset[key].label);
    const tensorsData = tensors
      .map((tensor, i) => {
        if (tensor) {
          const values = Object.keys(tensor).map(v => tensor[v]);
          return ml5.tf.tensor(values, dataset[i].shape, dataset[i].dtype);
        }
        return null;
      })
      .reduce((acc, cur, j) => {
        acc[j] = cur;
        return acc;
      }, {});
    knnClassifier.knnClassifier.setClassifierDataset(tensorsData);
  }
});

[oveddan]

Yes this would be great; with the json you could even save the data to the cloud on something like firebase.