Hands-On Python Deep Learning for the Web

Predicting using the TF.js model

After the training is done, we also need to create an API to invoke the prediction function and return the predicted result. We bind the API to the '/predict' route with a POST method to make a request to this API, as shown:

app.post('/predict', function(req, res) {    var test = tf.tensor2d([parseFloat(req.body.sepLen), parseFloat(req.body.sepWid),                                 parseFloat(req.body.petLen), parseFloat(req.body.petWid)], [1,4]);    var out = model.predict(test);    var maxIndex = 0;    for (let i=1;i<out.size; i++){        if (out.buffer().get(0, i) > out.buffer().get(0, maxIndex)){            maxIndex = i;        }    }        ans = "Undetermined";    switch(maxIndex) {        case 0:            ans = "Setosa";         break;        case 1:            ans = "Virginica";         break;        case 2:            ans = "Versicolor";  break; ...