Bird Species Embedding and Visualization using TIMM2Spotlight

This project processes a bird species dataset, extracts embeddings using a pretrained model, and visualizes the results using Renumics Spotlight.

Table of Contents

• Installation
• Download Dataset
• Usage
• Visualization

Installation

1. Clone the repository:

   git clone https://github.com/your-repo/bird-species-visualization.git
   cd bird-species-visualization

2. Create and activate a virtual environment. Install all requirements.

Download Dataset

1. Download the bird species dataset from Kaggle:

Go to the Kaggle dataset page and download the dataset: link

2. Extract the dataset:

Extract the downloaded dataset into the data directory within the project folder. The directory structure should look like this:

bird-species-visualization/
├── data/
│   ├── birds.csv
│   ├── test/
│   │   ├── birdClass1/
│   │   │   └── bird1.jpg...
│   │   └── birdClass2/...
│   ├── train/
│   └── valid/
├── dataset.py
├── spotlight_visualization.py
└── README.md

Usage

1. Run the script to process images and generate embeddings:

python spotlight_visualization.py

This script will:

• Load the bird species dataset.
• Use a pretrained model to generate embeddings for the images.
• Save the results to a CSV file.

2. Ensure the CSV file is created:

The script will create a file named bird_dataset_predictions.csv in the project directory.

Visualization

1. Visualize the results with Renumics Spotlight:

After running the script, the visualization will automatically launch in Renumics Spotlight, displaying the images and their embeddings.

Visualization Example