SPADE Semantic Image Synthesis Quiz

Answers and explanations appear as you take the quiz.

1. What input does SPADE use to generate an image?
   • A semantic segmentation map labeling each region
   • A spoken description
   • A single example photo
   • A 3D point cloud
2. What problem with earlier normalization did SPADE fix?
   • It made images too colorful
   • Normalization layers washed away the layout information
   • It used too little memory
   • It could not handle color images
3. In SPADE, the normalization scale and shift parameters are:
   • Random noise
   • Single values shared across the whole image
   • Spatially-varying tensors computed from the segmentation mask
   • Fixed constants
4. Which famous interactive app is built on SPADE?
   • Photoshop
   • Google Translate
   • Midjourney v1
   • GauGAN (NVIDIA Canvas)
5. What does the 'SP' in SPADE stand for?
   • Spatially-Adaptive
   • Super Pixel
   • Spectral Pooling
   • Sparse Projection
6. A practical benefit of SPADE's controllable layout is:
   • Improving the input map’s pixel resolution without image synthesis
   • Controlling where labeled scene elements appear in the generated image
   • Translating every label into natural-language captions
   • Compressing generated scenes into smaller files
7. Why can SPADE-generated images be useful when training semantic-segmentation systems?
   • The generated images contain only unlabeled audio
   • The system predicts a transcript for every pixel
   • The input layout already supplies pixel labels for the generated scene
   • The layout removes all semantic categories
8. What can happen when a designer changes a region label in a SPADE input map?
   • It changes the audio sample rate
   • It retrains the generator from scratch
   • It removes the labels from all other regions
   • The generator can synthesize the scene with the revised semantic region