Score
0
Streak
0
Accuracy
0%
Timer
00:00
Best
0
Level
1
Puzzle controls
PracticeOutput dimensions
Calculate the output tensor shape.
28 × 28 × 3
Conv 3 × 3
Feature-map visualisation
Input
Padding
Kernel
Output
Space pause or resume
N new puzzle
Enter submit
H hint
Formula and calculation
CNN pipeline builder
Parameters
0
Activations
0
Spatial reduction
0%
Session history
Solve a puzzle to create your history.
Level progress0 / 5
Layer dimensions and activations
Score and completion history
Learning guide
Convolution output
Apply padding first. Use the effective dilated kernel. Divide the usable span by stride, then apply the selected rounding rule.
Pooling output
Pooling usually preserves channels. Global pooling reduces each channel to one value, producing a 1 × 1 spatial map.
Common mistakes
Do not confuse filters with input channels. Include both padding sides, dilation gaps, and every spatial reduction before flattening.