Home▸AI & Machine Learning▸Multi-Layer CNN Forward Pass: Conv → Pool → Dense → Softmax (2D)

Multi-Layer CNN Forward Pass: Conv → Pool → Dense → Softmax (2D)

2D multi-layer CNN forward-pass lab: a drawable 12×12 input runs through two real multi-channel convolution layers, ReLU, max-pooling, flatten and a dense softmax classifier — every feature map computed live from real dot-product math, with adjustable kernel counts and reseedable weights.

AI & Machine Learning2DAdvanced60 FPS📱 Mobile-adapted⇄ 3D version

Details

CategoryAI & Machine Learning

RenderingCanvas 2D

DifficultyAdvanced

Target FPS60

ControlsPick or draw a 12×12 input pattern, adjust the Conv1/Conv2 kernel-count sliders, reseed the random weights, and step through all nine pipeline stages (Input → Conv1 → ReLU1 → Pool1 → Conv2 → ReLU2 → Pool2 → Flatten → Softmax) manually or with auto-advance.

convolutional neural networkmulti-channel convolutionmax poolingrelu activationsoftmax classifierforward propagation

Related simulations

Convolutional Neural NetworkThe 3D version of this simulation.

2D Image Convolution: Edge Detection & KernelsA single-kernel convolution lab — this sim instead runs a full multi-layer network with pooling, flatten and softmax. ⊞ All in AI & Machine Learning →

This 2D companion runs a genuine small convolutional neural network end to end: a drawable 12×12 input passes through a real multi-kernel Conv1 layer, ReLU, 2×2 max-pooling, a real multi-channel Conv2 layer that mixes every Pool1 channel, another ReLU and pool, a flatten step and a dense layer whose logits are turned into softmax class probabilities — every one of the nine pipeline stages is real matrix math computed live, not a single-kernel demo.

⚙ Under the hood

2D multi-layer CNN forward-pass lab with two real multi-channel convolution layers, ReLU, max-pooling, flatten and a dense softmax classifier, all recomputed live from adjustable kernel counts and reseedable random weights.

convolutional neural networkmulti-channel convolutionmax poolingrelu activationsoftmax classifierforward propagation

2D · HTML5 Canvas 2D · 60 FPS target · runs fully client-side, no install