Project

Automating Port Operations

A vessel-classification project built around image understanding, transfer learning, deployment tradeoffs, and live browser demonstration.

Project at a glance

This page presents the vessel classifier, the notebook that produced it, and the saved outputs that support the published result.

Run in Colab View Notebook Open Metrics JSON

The goal is to classify vessels from images using the published notebook workflow and the saved model comparison outputs.

The final result favors transfer learning, and the page keeps the notebook, evidence, and metrics easy to open from one place.

The requirement list that drives the notebook sections.

Saved notebooks, plots, metrics, and supporting notes.

Baseline loss and accuracy history.

Class-by-class review heatmap for the baseline lane.

MobileNetV2 training history for the winning lane.

Saved model comparison and held-out accuracy values.

The notebook note that records why transfer learning is preferred.

Notebook access

Open the notebook source or launch it in Colab to rerun the workflow.

Run in Colab View Notebook

Transfer learning is the published recommendation.

The comparison summary records 84.33% held-out accuracy for transfer learning versus 44.30% for the custom CNN.

The observation text records why transfer learning is preferred and keeps the project decision explicit.