File size: 3,363 Bytes
e418977 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 |
## Image Similarity Search Engine
A deep learning-based image similarity search engine that uses EfficientNetB0 for feature extraction and FAISS for fast similarity search. The application provides a web interface built with Streamlit for easy interaction.
Features
- Deep Feature Extraction: Uses EfficientNetB0 (pre-trained on ImageNet) to extract meaningful features from images
- Fast Similarity Search: Implements FAISS for efficient nearest-neighbor search
- Interactive Web Interface: User-friendly interface built with Streamlit
- Real-time Processing: Shows progress and time estimates during feature extraction
- Scalable Architecture: Designed to handle large image datasets efficiently
## Installation
## Prerequisites
Python 3.8 or higher
pip package manager
## Setup
1. Clone the repository:
```
git clone https://github.com/yourusername/image-similarity-search.git
cd image-similarity-search
```
2. Create and activate a virtual environment:
```
python -m venv venv
source venv/bin/activate # On Windows use: venv\Scripts\activate
```
3. Install required packages:
```
pip install -r requirements.txt
```
## Project Structure
```
image-similarity-search/
├── data/
│ ├── images/ # Directory for train dataset images
│ ├── sample-test-images/ # Directory for test dataset images
│ └── embeddings.pkl # Pre-computed image embeddings
├── src/
│ ├── feature_extractor.py # EfficientNetB0 feature extraction
│ ├── preprocessing.py # Image preprocessing and embedding computation
│ ├── similarity_search.py # FAISS-based similarity search
│ └── main.py # Streamlit web interface
├── requirements.txt
├── README.md
└── .gitignore
```
## Usage
1. **Prepare Your Dataset:**
Get training image dataset from drive:
```
https://drive.google.com/file/d/1U2PljA7NE57jcSSzPs21ZurdIPXdYZtN/view?usp=drive_link
```
Place your image dataset in the data/images directory
Supported formats: JPG, JPEG, PNG
2. **Generate Embeddings:**
```
python -m src.preprocessing
```
**This will**:
- Process all images in the dataset
- Show progress and time estimates
- Save embeddings to data/embeddings.pkl
3. **Run the Web Interface:**
```
streamlit run src/main.py
```
4. Using the Interface:
- Upload a query image using the file uploader
- Click "Search Similar Images"
- View the most similar images from your dataset
## Technical Details
**Feature Extraction**
- Uses EfficientNetB0 without top layers
- Input image size: 224x224 pixels
- Output feature dimension: 1280
**Similarity Search**
- Uses FAISS IndexFlatL2 for L2 distance-based search
- Returns top-k most similar images (default k=5)
**Web Interface**
- Responsive design with Streamlit
- Displays original and similar images with similarity scores
- Progress tracking during processing
**Dependencies**
- TensorFlow 2.x
- FAISS-cpu (or FAISS-gpu for GPU support)
- Streamlit
- Pillow
- NumPy
- tqdm
**Performance**
- Feature extraction: ~1 second per image on CPU
- Similarity search: Near real-time for datasets up to 100k images
- Memory usage depends on dataset size (approximately 5KB per image embedding)
|