Overview

Example of deploying a pre-trained serialized Tensorflow model trained on the MNIST data set to predict hand written digits.

Usage

Note for this algorithm to run you must have the MNIST image data in your own data collection in the following format or simply use the runnable test example.

Input

{"mnist_images": "data://YOUR_USERNAME/tensorflow_mnist_data/t10k-images-idx3-ubyte.gz", "mnist_labels": "data://YOUR_USERNAME/tensorflow_mnist_data/t10k-labels-idx1-ubyte.gz"}

Output

{"prediction": prediction, "average accuracy": accuracy}

Examples

Input:

{"mnist_images": "data://demo/tensorflow_mnist_data/t10k-images-idx3-ubyte.gz", "mnist_labels": "data://demo/tensorflow_mnist_data/t10k-labels-idx1-ubyte.gz"}

Output:

"MNIST Predictions: {'prediction': array([7, 2, 1, ..., 4, 8, 6]), 'accuracy': 0.8891}, TF version: 1.10.0"

demo / TensorflowDemoGPU / 0.1.1

Comments

Loading comments