Quick Start Guide

Installation

Option 1: Download Release Binary

Download the latest pre-built binary for your platform:

unzip miniAI-linux-x64.zip
cd miniAI
./miniAI help

Option 2: Build from Source

Prerequisites: GCC (or Clang on macOS), Make, Python 3 (optional, for PNG generation)

git clone https://github.com/nelsonramosua/miniAI.git
cd miniAI
make
./miniAI help

macOS note: OpenMP is handled automatically by the Makefile. If you hit issues, install libomp:

brew install libomp

Option 3: Docker

docker pull nelsonramosua/miniai:latest
docker run --rm nelsonramosua/miniai help

Your First Model

Train on digits (fast, ~5 seconds)

./miniAI train --dataset digits --static

Test the model

./miniAI test --dataset digits --static

That’s it — you trained a neural network in C!

Common Workflows

Static workflow (fast, for development)

# Train
./miniAI train --dataset digits --static

# Test
./miniAI test --dataset digits --static

# Find best hyperparameters
./miniAI benchmark --dataset digits --static --reps 3

# Test again with optimized config
./miniAI test --dataset digits --static

PNG workflow (realistic images)

# Train
./miniAI train --dataset digits --data

# Test on dataset
./miniAI test --dataset digits --data

# Test on a single image
./miniAI test --image my_digit.png

# Benchmark
./miniAI benchmark --dataset digits --data --reps 3

Alphanumeric + phrase recognition

# Train alphanumeric PNG model
./miniAI train --dataset alpha --data

# Recognize a phrase in an image
./miniAI recognize --image IO/images/testPhrases/hello_world_train.png