1. When you train an image-recognition model on billions of images and hashtags, what will cause the performance of the model to increase?

A. Adding more layers to the model

B. Adding more parameters in each layer of the model

C. Adding more images to the training data

D. All of the above

2. What is the objective in semantic segmentation?

A. Provide a class prediction for every pixel in the image

B. Provide a class prediction for the image

C. Provide bounding boxes for objects in the scene, each with a class prediction.

D. Provide a natural text description of the image contents

3. In a real-world deployment of an image-recognition model, what is typically the most important?

A. The model can be trained very quickly

B. The model makes predictions very quickly

C. The model has as many layers as possible

D. The model is trained on many GPUs

4. In a multi-scale DenseNet model for image recognition, what does the "scale" refer to?

A. The scale of the input image

B. The number of parameters in the model

C. The scale of the feature maps

D. The number of compute operations required to make a prediction