Classification of alcohol brand logos using convolutional neural networks
Pichitchai Pimkote, Thanapat Kangkachit · 2018
In Thailand, intentionally posting pictures with alcohol logos in social media is a violation of the Alcohol Control Act and should face tough legal action by police officer. Thailand also remains in top countries for social media. Hence, we need an automated approach for classification of alcohol brand logos in large numbers of pictures posting in social media. Recently, convolutional neural networks (CNNs) have been widely adopted to recognize logos in images. In this work, a CNNs model is built to classify whether the input image is alcohol or non-alcohol logo. In case of classifying as alcohol logo, its corresponded alcohol brand is provided. Experimental results show that our CNNs model yields better results in terms of precision and recall compared to the all-in-one CNNs model.