Hi, I’m confused by the description regarding finding the class of a detected object in the assignment. As the first screenshot shows, isn’t [c1, c2, c3] just a one-hot encoded vector containing only 1 and 0? Why does c_i then become the probability that the object is the class i?
You are confusing probabilities outputs of the model with label encodings. In this case the c is the class probability output for class c3 which can be one hot encoded or numeric!
1 Like