Mastering Computer Vision with TensorFlow 2.x

Caution when using the neural network model

Note that we only used 68 images to train our neural network and it gave us very good predictions. This brings up four questions:

1. Does the model that we developed predict correctly in all situations? The answer is no. The model has only two classes—burger and fries—so it may detect other objects that are similar, for example, to the shape of a burger, such as a donut. To fix this issue, we need to load images similar to burger and classify them as not burger, then train the model with those additional image sets.
2. Why do we hear that we need thousands of images to train a neural network? If you are training a neural network from scratch, or even using transfer learning to get weights from another ...