CS5720 - Week 8
Slide 150 of 160
Interactive Transfer Learning Demo
Strategy:
Feature Extraction
Fine-Tuning
Progressive
Start Demo
Reset
Compare All
Network Architecture
Conv Block 1
64 filters
Conv Block 2
128 filters
Conv Block 3
256 filters
Conv Block 4
512 filters
Classifier
New Task
Legend:
🔒 Frozen
|
🔓 Trainable
|
✨ New
Training Metrics
0%
Accuracy
-
Loss
0s
Training Time
0
Epochs
Current Strategy:
Feature Extraction - Training only new classifier layers
Strategy Comparison
Training from Scratch
Final Accuracy:
72%
Training Time:
8 hours
Data Required:
50,000+
Computational Cost:
High
Feature Extraction
Final Accuracy:
89%
Training Time:
30 minutes
Data Required:
5,000+
Computational Cost:
Low
Fine-Tuning
Final Accuracy:
94%
Training Time:
2 hours
Data Required:
10,000+
Computational Cost:
Medium
Click on each approach
to explore detailed analysis and use cases
← Previous
Next →
Prepared by Dr. Gorkem Kar
Modal Title
×
Modal content goes here...