Katy Or Zooey? Image Search and Classification with DuckDuckGo and FastAI

Katy Perry or Zooey Deschanel detector

DEMO

You can try the code in Hugging Face Spaces

Link: https://huggingface.co/spaces/xota1999/Katy_Perry_or_Zooey_Deschanel

Project Overview

This project demonstrates how to leverage the DuckDuckGo search engine to gather images, create a dataset, and train an image classification model using the FastAI library. The example focuses on downloading images of celebrities and using these images to train a simple neural network for classification tasks.

Features

• Image Search: Use DuckDuckGo to search and download images based on specified search terms.
• Dataset Creation: Organize downloaded images into a structured dataset.
• Image Verification: Verify the integrity of images and remove any corrupted files.
• Model Training: Train an image classification model using FastAI's high-level API.
• Model Evaluation: Test the model's performance on new images and save the trained model for future use.

Installation

To get started, install the required dependencies:

pip install -U duckduckgo_search fastcore fastdownload fastai

Usage

1. Search and Download Images:

   • Define search terms and download images using DuckDuckGo.
   • Save the images locally for dataset creation.

2. Create Dataset:

   • Organize the downloaded images into a dataset structure suitable for training.
   • Perform necessary image preprocessing and resizing.

3. Train the Model:

   • Use FastAI to create data loaders and define a convolutional neural network (CNN) model.
   • Train the model on the dataset and fine-tune it to improve performance.

4. Evaluate the Model:

   • Test the model on new images to assess its accuracy.
   • Save the trained model for deployment or further use.

License

This project is licensed under the MIT License. See the LICENSE file for more details.