Draw Grid
Load a model (.onnx) to begin.
Neural Network Visualizer
Draw a digit or letter and watch activations/probabilities update in real time.
Layer/Tensor Interaction
Output Layer (0-9)
Prediction: -
Draw Grid
Switch to this tab to auto-load the character model.
Layer/Tensor Interaction
Output Layer (A-Z)
Prediction: -
About This Project
This project is an interactive MNIST digit and EMNIST character recognition demo that runs entirely in the browser using ONNX Runtime Web. You can draw on an exact 28x28 input grid, see live prediction probabilities for digits 0-9 or letters A-Z, and visualize model-layer interactions in real time.
Created by Aditya Agarwal, focused on practical machine learning interfaces, AI product engineering, and explainable model behavior.
Source code: github.com/adiagarwalrock/mnist-digits-web