README.md

LogoDet

Neural Nets for logo detection

Using the free API

• Signup here for free API key.
• Maximum 16 requests per hour allowed.
• Each request can have maximum of 4 images to be processed. (i.e: maximum of 64 images can be processed per hour.)

Example usage of the free API

wget https://github.com/notAI-tech/fastDeploy/blob/master/cli/fastDeploy-file_client.py
chmod +x fastDeploy-file_client.py

# with webhook example
./fastDeploy-file_client.py --file ../../LogoDet/example.jpg  --url "https://tech.notai.tech/logodet/async?api_key=YOUR_API_KEY" --webhook https://fastdeploy.requestcatcher.com

# without webhook example
./fastDeploy-file_client.py --file ../../LogoDet/example.jpg  --url "https://tech.notai.tech/logodet/async?api_key=YOUR_API_KEY"

As a service (recommended)

# Start service
(sudo) docker run --name logodet -p 8080:8080 notaitech/fastdeploy-recipe:logodet

# Running predictions
wget https://github.com/notAI-tech/fastDeploy/blob/master/cli/fastDeploy-file_client.py
chmod +x fastDeploy-file_client.py

# Single input
./fastDeploy-file_client.py --file PATH_TO_YOUR_IMAGE

# Client side batching
./fastDeploy-file_client.py --dir PATH_TO_FOLDER --ext jpg

As a Python module

git clone https://github.com/notAI-tech/LogoDet/
cd LogoDet
pip install -r requirements.txt

# Weights will be auto-downloaded
from predictor import predictor

image_paths = [image_1.jpg, image_2.png, ...]

predictions = predictor(image_paths)

Notes:

1. Although we were able to generate/ gather data for more than 2000 unique company logos, the current release is limited to these 292 logos due to hardware constraints.
2. We recommend using LogoDet via fastDeploy, as it doesn't require the user to install dependencies separately.