How to get bounding box from custom trained model detection?

Hi,

I’ve trained a custom model with one object, converted it into graph model and using on the browser. Here is my code:

        async function predict() {
            const model = await tf.loadGraphModel('./model/model.json');
            let img = document.getElementById('test');

            var example = tf.browser.fromPixels(img);
            example = example.expandDims(0);

            const output = await model.executeAsync(example);
            output.forEach(val=>{
                console.log(val);
            });
        }

It’s giving these values:

t {kept: false, isDisposedInternal: false, shape: Array(3), dtype: "float32", size: 1200, …}
t {kept: false, isDisposedInternal: false, shape: Array(3), dtype: "float32", size: 600, …}
t {kept: false, isDisposedInternal: false, shape: Array(1), dtype: "float32", size: 1, …}
t {kept: false, isDisposedInternal: false, shape: Array(3), dtype: "float32", size: 600, …}
t {kept: false, isDisposedInternal: false, shape: Array(3), dtype: "float32", size: 1200, …}
t {kept: false, isDisposedInternal: false, shape: Array(5), dtype: "float32", size: 4915200, …}
t {kept: false, isDisposedInternal: false, shape: Array(2), dtype: "float32", size: 300, …}
t {kept: false, isDisposedInternal: false, shape: Array(2), dtype: "float32", size: 300, …}

How can I know is there any mathces/predictions on the image? And if is there, how can I get the bounding box of it?

The pretrained models have these options. But how can we do it on our own models?

[tafsiri]

I’m going to close this issue since the original problem has been solved. But if you are looking for more general “how can I do X” tips. I would suggest using StackOverFlow, there is both a tensorflow tag and a tensorflow.js that may have some helpful pointers. You might also find some of the online courses listed on this page helpful.

[sundowatch]

I’ve now trained with AutoML and it works fine. But don’t know why can’t I use it with tensorflow training.