Dog Search

This project's origin is here.

Description

In this demo, we will use Weaviate to create an image-search application. We'll develop a dog image-search application that lets users upload a photo of a dog and receive a list of dog breeds that are the most similar.

(TODO: Add demo video)

Used Technologies

• Weaviate Vector DB
• Python Programming Language (v3.10+)
• Weaviate Python Client Library
• Flask (For WebApp interface development)
• Docker Desktop

Weaviate Modules

Image Vectorization

In this demonstration, we will work with a dataset of dog images. The current dataset includes ten images of different dog breeds. However, you can easily replace these images with any other dataset to suit your specific use case. To do this, add your images to the flask-app/static/img folder and execute the images-to-base64.py and upload-data-objects.py files.

For this use case, it's essential that image vectorization captures information about the dogs, such as breed, size, and color. Weaviate's img2vec-neural module is designed to solve this problem. It vectorizes each image, enabling semantic similarity-based searches.

Img2vec-neural Module

Weaviate's img2vec-neural module is a flexible vectorizer that converts images into meaningful vectors. It supports the ResNet-50 model, a Convolutional Neural Network (CNN) trained on the ImageNet database, which contains over 10 million images and 20,000 classes.

Prerequisites

Before you begin, ensure you have met the following requirements:

• Docker Desktop should be installed and running.
• Python 3.10 or later should be installed.

Setup Instructions

Note: We recommend you create a new virtual environment for this.

python -m venv .venv
source .venv/bin/activate

1. Install the Python packages.

pip install -r requirements.txt

2. Run the Docker file.

docker compose up

3. Create the class definition.

python create-collection.py

4. Convert images to base64 encodings.

python images-to-base64.py

5. Upload the encoded images.

python upload-data-objects.py

Usage Instructions

Run the application.

python flask-app/application.py

Access the application by opening your web browser and visiting http://localhost:5001.

Run the query

Run the query to see dogs that are under 60 pounds.

python query.py

Dataset License

The dataset currently contains ten images of different dog breeds. You can also build on this and add your own images to the dataset!