Inference Robot

The goal of this project is to initiate an inference project using Nvidia DIGITS workspace to build classifier depending on Transfer learning like ( LeNet ,AlexNet , googleNet ) Architectures.

Project divide into two Tasks:

• First part is an inference project against supplied digit image recognition dataset

• Second part is to build inference project from the data collected by me. For more information see my pdf.

Result

Classified hand-written digits images on LeNet network model with a best accuracy of 90.19 % and an inference time of less than 10 ms.

Classifying digits numbers using LeNET

Classified sort of cups images based on AlexNet and GoogleNet networks model with a best accuracy of 99.17 % .

Classifying Mug using AlexNet

Classifying Cup of Coffee using GoogleNet