keras-model-client

Python client API for the Keras Model Server.

Install from PyPI:

pip install keras-model-client

Basic usage:

from keras_model_client import predict

# Using Detectron2 with the default model server
result = predict('/path/to/image.jpg', 'image_queue_detectron2')

# Using YOLOv5 with a custom model server
result = predict('/path/to/image.jpg', 'image_queue_yolov5', url='http://your-model-server/predictor/')

print(result)
# PredictionResult(success=True, predictions=[...])

Note: The default model server is currently hosted at https://adamant.tator.io:8082/.

CLI

The package provides a command-line interface invoked with model-server-predict. For example:

model-server-predict /path/to/image.jpg image_queue_detectron2 -f json

will run the image at /path/to/image.jpg through the Detectron2 model and format the results as JSON. Additional help is available with the -h or --help flag.

Install from source

Clone the repository:

git clone https://github.com/kevinsbarnard/keras-model-client
cd keras-model-client

and install from source with Poetry:

poetry install

or with pip:

pip install .