Images are made by taking a photo or created by artists and amature hobbyists drawing on paper. Can computer classify that correctly, forget accuracy for a while. Can it even read the image? That is where P5, Javascript comes to rescue. We will see, how it does
Can the browser identify you? With ML5 to support, the answer is Yes. It can not just identify you, it can go beyong and give added information on how it identified you.
The live Classification must have shown you that real images with multiple objects inside a picture can create confusing predictions. When the scene is seperated in to objects, and when it is detected then you will see more accuracy.
Models are JSON files looking up the mathematical weights and biases to provide the prediction. The mobilenet model is downloaded from the cloud, but we can Roll our Own model.