Project Ecosystem

This repository is part of a three-tier fish identification system:

• Deep Learning Model - Core Deep learning model for fish species identification and Similarity Search.
• API Backend - RESTful API serving the Deep learning model
• AquaVision Frontend - Flutter Application for fish identification

Marine Drive Biodiversity Monitoring Using Deep Learning

This project focuses on monitoring marine biodiversity by detecting and classifying fish species using deep learning models.
It uses cropped fish images and applies transfer learning (EfficientNet) to classify species effectively.

---

📊 Dataset:

• https://www.kaggle.com/datasets/markdaniellampa/fish-dataset
• Format: Cropped .png images
• Structure: Images are named with species labels.
• The notebook extracts species names from filenames and prepares them for training/validation.

---

⚙️ Requirements

• Python 3.x
• Jupyter Notebook
• torch, torchvision
• PIL
• seaborn, matplotlib
• pyngrok (for Colab sharing)
• google.colab (if running in Google Colab)

Install requirements (if not using Colab):

pip install torch torchvision pillow seaborn matplotlib pyngrok

### 📈 Results

Trained an EfficientNet-B0 model on cropped fish images.

Accuracy trends and performance visualizations provided.

Species-wise classification analysis included.