Skip to content

Latest commit

 

History

History
24 lines (19 loc) · 891 Bytes

README.md

File metadata and controls

24 lines (19 loc) · 891 Bytes

bottle-image-classifier

Image Classification with LeNet and AlexNet

Overview

A 5-class liquid amount classification task.

Dataset

Background information

The bottle dataset was collected during ECBM E4040 Fall 2016. There are 2 types of bottles: coke bottle and water bottle, and the students were asked to take pictures with their cellphones. A post-processing was done to make sure each picture has the same size.

There are 5 classes in total:

0% (labeled as 0) 25% (labeled as 1) 50% (labeled as 2) 75% (labeled as 3) 100% (labeled as 4) All those labels are visual estimation of the actual amount.

Data organization

Training set: 15000 images in total, all the classes are balanced(3000 images per class). Data available on kaggle. Test set 3500 images in total

Models

I implemented a modified LeNet and AlexNet with 85% and 91% accuracy on test data respectively