concept detection in category deep learning

This is an excerpt from Manning's book Deep Learning with JavaScript: Neural networks in TensorFlow.js.
iVarun Gulshan et al., “Development and Validation of a Deep Learning Algorithm for Detection of Diabetic Retinopathy in Retinal Fundus Photographs,” JAMA, vol. 316, no. 22, 2016, pp. 2402–2410, http://mng.bz/wlDQ.
Listing 3.5. Defining a binary-classification model for phishing detection (from index.js)
const model = tf.sequential(); model.add(tf.layers.dense({ inputShape: [data.numFeatures], units: 100, activation: 'sigmoid' })); model.add(tf.layers.dense({units: 100, activation: 'sigmoid'})); model.add(tf.layers.dense({units: 1, activation: 'sigmoid'})); model.compile({ optimizer: 'adam', loss: 'binaryCrossentropy', metrics: ['accuracy'] });
The new task we will use to illustrate this point is object detection, the first nonclassification computer-vision problem type you encounter in this book. Object detection involves detecting certain classes of objects in an image. How is it different from classification? In object detection, the detected object is reported in terms of not only its class (what type of object it is) but also some additional information regarding the location of the object inside the image (where the object is). The latter is a piece of information that a mere classifier doesn’t provide. For example, in a typical object-detection system used by self-driving cars, a frame of input image is analyzed so that the system outputs not only the types of interesting objects that are present in the image (such as vehicles and pedestrians) but also the location, apparent size, and pose of such objects within the image’s coordinate system.