Tools & Assets
· 2019
Digit Recognition
Unity | C# | neural network from scratch
A handwritten-digit recognizer built on a feed-forward neural network I wrote from scratch in C#, without any ML libraries, and sold on the Unity Asset Store as a learning resource years before LLMs made neural networks a household topic. It shipped with two demos, digit recognition and XOR learning, plus a real-time visualization of the network's weights and activations that runs with almost no performance cost.
Highlights
- Feed-forward network with backpropagation, implemented from scratch in C# with no external ML libraries.
- Real-time network visualization: weights and activations update live during training.
- Two demos: handwritten digit recognition and XOR learning.
- Sold on the Asset Store as an educational template (now retired).
