MLX Fish Species Classifier
A custom CNN for fish species recognition, trained and deployed locally on Apple Silicon using MLX.
What Is It?
This project is an end-to-end, hardware-accelerated deep learning pipeline built from scratch using Apple's MLX framework. This project trains a custom VGG-style Convolutional Neural Network (CNN) to classify 9 distinct species of fish with extremely high accuracy.
By leveraging MLX, the model takes full advantage of unified memory architecture on Apple Silicon (M-series chips). This allows for near-instant memory access between the CPU and GPU without the traditional PCIe bottleneck, enabling extremely fast training and inference directly on a MacBook without cloud compute.
The Experiment
The project's true story isn't just about training a CNN—it's about overcoming data challenges and fighting overfitting.