Project 3 Q&A: Jonathan Krause
Project 3 Q&A: Jonathan Krause
Jonathan Krause
region
region no context
16 pixels
works best
Girshick et al., 2014
• Class-specific
• Ridge regression on bounding box offset
(cx, cy, log(width), log(height))
• Regularization amount super important
• Larger CNN
• AlexNet or VGGNet?
• Fine-tuning
• How much does it help in this case?
• Segmentation
• Combine project 1 and 3
• Multi-label prediction
• Predict attributes of objects?
• Make it better
• Add some other signal???
• Analyze it
• What really makes R-CNN tick?