ImageNet Explanation

Course Q&A Deep Learning Specialization Convolutional Neural Networks
1

in the assignment 2 of the week3 , when we used the imageNet pretrained model before we apply transfer learning and fine tuning to it , when we gave the model a bunch of images of Alpaca ,the
output was a wrong labels from the model classes ,but it should be like “Unknown object/image” , because if the model
trained to classify a 10 classes the model should has a 11 classes the last one is “Unknown”
Is That right ?

Thank u

2

Not true.

The model picks the best match from the classes it has learned.

Adding an “unknown” class requires some additional decision data.