🐟 Marine Species Classifier using Deep Learning and Computer Vision

A web-based deep learning application that classifies underwater marine species from images and visualizes model attention using Grad-CAM.

---

Demo

Demo

🌐 Live Demo

▶️ Deployed on Streamlit Cloud:
🔗 marine-species-classificationapp.streamlit.app

---

🧠 About the Project

This project applies deep learning and computer vision to automatically classify different types of marine species from underwater images.

It also includes Grad-CAM (Gradient-weighted Class Activation Mapping) to interpret what the model is focusing on while making predictions.

---

📂 Dataset

Dataset Source: Fish Species Image Dataset (Kaggle)

• 💾 ~4,000 cropped images of various marine species
• 📁 Images stored in /Fish_Data/images/cropped/
• ✅ Dataset used to train and evaluate a ResNet-based CNN classifier

---

🔧 Tech Stack

Component	Tool
Language	Python 3.12
Deep Learning	PyTorch, torchvision
Web App	Streamlit
Visualization	Matplotlib, OpenCV
Deployment	Streamlit Cloud

---

🧪 Model Info

• 📚 Base Model: ResNet18
• 🔄 Fine-tuned on marine species dataset
• 🧠 Last fully connected layer adjusted for custom class count
• 🧾 Saved in float16 precision for faster loading

---

🚀 Features

• 📤 Upload any marine species image
• 🤖 Get instant top predictions with confidence scores
• 🔍 View Grad-CAM heatmaps to understand model focus
• 📊 Simple, interactive UI

---

📁 Project Structure

marine-species-classification.app/ ├── app.py
├── resnet18_half_precision.pth
├── Marine_Species_Classifier_using_Deep_Learning.ipynb
├── requirements.txt
└── README.md

👨‍💻 Author Ashutosh Kumar 🎓 IIT Patna 📧 ashutosh_2312res778@iitp.ac.in 🔗 LinkedIn | GitHub

🛠️ Local Setup Instructions

# Clone the repo
git clone https://github.com/ashutosh8021/marine-species-classification.git
cd marine-species-classification

# Install dependencies
pip install -r requirements.txt

# Run the Streamlit app
streamlit run app.py