Published paper:

Resource-aware On-device Deep Learning for Supermarket Hazard Detection

Arxiv

Description

This repository contains the necessary code and instructions to build and fine-tune deep learning(DL) model for grocery hazard detection. We used MXNet framework to build, train and deploy all the deep learning models.

1. Fine-tune existing DL models
2. Build a new model and fine-tune it

Prerequisite

1. Install MXNet

Test Results

Here are some test results. The Azure color box at the bottom of the picture represents the model's confidence for a particular image.