Level 5 · Coming soon
CNN Lab
Convolutional networks read images the way earlier levels read numbers and words. You'll see filters slide across a handwritten digit, watch feature maps light up, and train a tiny CNN on an MNIST-style dataset.
Planned subtasks
- Slide a filter — pick a 3×3 kernel, watch it detect edges/corners.
- Read the feature map — guess which digit lit up which layer.
- Train tiny — go small — beat a target accuracy on MNIST with the fewest neurons.