Arminkhayati / visual_search_engine

A YOLO v3 based visual search engine.

Visual Search Engine

A visual search engine powered by YOLO v3 and fastapi. In this project, a text-based description is received from users and the images that match that description are retrieved and displayed in the order of priority. That is what we do with Google to find the images we want but in more simple manner in our case.

Code uses tensorflow 2.x and the base model is taken from this repository.

Installation

First, clone or download this GitHub repository:

git clone https://github.com/Arminkhayati/visual_search_engine.git
cd visual_search_engine

then prepare the environment:

pip install -r ./requirements.txt

or

conda env create -f environment.yml

Then download weights and place them in this directory vse/model/model_data:

# yolov3
wget https://pjreddie.com/media/files/yolov3.weights

# yolov3-tiny
wget https://pjreddie.com/media/files/yolov3-tiny.weights

Running code

To run the server just activate your environment then run this command:

For testing:

uvicorn main:app --reload

For mannually running server:

uvicorn main:app --host 0.0.0.0 --port 80