Efficient CNN Models for Beer Bottle Cap Classification Problem

Abstract

In this work, we present an efficient solution to the beer bottle cap classification problem. This problem arises in the Wecheer smart opener project. Although classification problem is common in Computer Vision, there is no dedicated work for beer bottle cap dataset. We combine state-of-the-art deep learning techniques to solve the problem. Our solution outperforms the well-known commercial system that is currently used by the Wecheer project. It is also more efficient than the famous architectures such as VGG, ResNet, and DenseNet for our purposes.