🐱🐶 Image Tagging System using TensorFlow & Streamlit

An end-to-end Image Tagging (Image Classification) system built using TensorFlow (CNN + Transfer Learning) and deployed with a Streamlit frontend.
The application allows users to upload an image and receive a predicted label along with a confidence score.

---

📌 Project Overview

This project implements an image tagging system that classifies uploaded images into predefined categories such as Cat and Dog.
A pre-trained MobileNetV2 model is used for feature extraction, making the system CPU-efficient and suitable for execution on local machines.

To improve real-world reliability, the application includes confidence-based unknown detection, preventing incorrect predictions for unrelated images (for example, human faces).

---

🎯 Objectives

• Develop an image classification model using Convolutional Neural Networks (CNN)
• Apply transfer learning to reduce training time and CPU usage
• Preprocess and normalize image data
• Deploy the trained model using a frontend web interface
• Display prediction results with confidence scores
• Handle unknown or unrelated images intelligently

---

🧠 Technologies Used

• Programming Language: Python
• Deep Learning Framework: TensorFlow & Keras
• Pre-trained Model: MobileNetV2
• Frontend Framework: Streamlit
• Image Processing: Pillow (PIL), NumPy
• Version Control: Git & GitHub

---

🏗️ System Architecture

User Uploads Image
↓
Streamlit Frontend (frontend/app.py)
↓
Image Preprocessing (Resize & Normalize)
↓
MobileNetV2 (Feature Extraction)
↓
Custom CNN Classification Layer
↓
Softmax Output (Probabilities)
↓
Prediction + Confidence / Unknown Handling

---

📁 Project Folder Structure

Image_Tagging_Project/
├── dataset/
│   ├── cat/
│   │   ├── cat1.jpg
│   │   ├── cat2.jpg
│   │   └── ...
│   └── dog/
│       ├── dog1.jpg
│       ├── dog2.jpg
│       └── ...
├── model/
│   └── image_tagger.h5
├── frontend/
│   └── app.py
├── train.py
├── predict.py
├── requirements.txt
├── .gitignore
└── README.md

---

⚙️ How the System Works

1. Images are stored in folders where each folder name represents a class label.
2. Images are resized to 128 × 128 pixels and normalized.
3. MobileNetV2 extracts high-level image features.
4. A custom dense layer predicts the image class.
5. The Streamlit frontend allows users to upload images.
6. The model outputs prediction probabilities using Softmax.
7. If confidence is below a defined threshold, the image is classified as Unknown.

---

🧪 Model Details

• Model Type: Convolutional Neural Network (CNN)
• Learning Approach: Transfer Learning
• Base Model: MobileNetV2 (pre-trained on ImageNet)
• Input Shape: 128 × 128 × 3
• Optimizer: Adam
• Loss Function: Categorical Crossentropy
• Epochs: 3 (CPU-friendly training)

---

🖥️ Frontend Features

• Upload image via browser
• Display uploaded image
• Predict image category
• Show confidence percentage
• Handle unknown images gracefully
• Simple and user-friendly interface

---

🚀 How to Run the Project

1️⃣ Install Dependencies

pip install -r requirements.txt

2️⃣ Train the Model

python train.py

This will generate:

model/image_tagger.h5

3️⃣ Run the Frontend Application

cd frontend
streamlit run app.py

Open in browser:

http://localhost:8501

⚠️ Unknown Image Handling

If the prediction confidence is below 70%, the system displays:

⚠️ Unknown object (not cat or dog)

📊 Sample Output

Prediction: CAT
Confidence: 95.91%

OR

Unknown object (not cat or dog)
Confidence was only 48.20%

📈 Future Enhancements

• Add more image classes (human, car, animal types)
• Implement multi-label image tagging
• Deploy using Flask or FastAPI
• Add cloud deployment (AWS / GCP)
• Improve UI with custom CSS
• Integrate real-time camera input

---

🎓 Learning Outcomes

• Understanding of CNNs and transfer learning
• Practical experience with TensorFlow & Keras
• Frontend integration using Streamlit
• Real-world handling of unknown data
• GitHub project structuring and version control

---

📜 License

This project is developed for educational purposes and is free to use for learning and experimentation.

---

👤 Author

Dikshitha A
Aspiring Software Engineer | AI & ML Enthusiast | Python Development