Hey, Jetson!

Automatic Speech Recognition Inference on the Nvidia Jetson.

Includes: Python, Flask, Tensorflow, Keras, Nvidia Jetson, Jetpack, Ubuntu, L4T, HTML, CSS, JavaScript, Anaconda, Jupyter Notebook, Cudnn

Introduction

This server runs a live demo of a scalable automatic speech recognition platform in Keras/Tensorflow for inference on the Nvidia Jetson Embedded Computing Platform for AI at the Edge. This end to end sequencing model uses a recurrent neural network trained on an Nvidia GTX1070. The overall cosine similarity of the model's predictions with the ground truth transcriptions in both the test and validation set is about 74%, while the overall word error rate is about 18%. Use the inference engine below to test out the model.

Inference Engine

Select an instance to run inference on. Valid indexes are:
Train set: 0-25150
Validation set: 0-2022
Test set: 0-1878