Incognition logo
Home
Learn
Practice
Leaderboard
0
VisualizationsDeep Learning for Images

VGG Network

Click to swap one 5×5 filter for two 3×3 filters in a row. Both see a 5×5 patch, but the pair needs 18 weights, not 25, so VGG uses only 3×3.

region seen
5×5
weights
25

Same view, fewer weights, and an activation in between. VGG took this to its conclusion: use nothing but 3×3.

Customize