Spam detection in emails has been a standard classification problem since decades. But nowadays, emails are only a small part of our digital conversations. Instant messengers, for example, are a huge source for spam messages and images these days. This paper proposes an application of deep CNN’s to help classify these images into three main classes: Important, Acceptable and Spam. Acceptable are those images which might be irritating, but are not necessarily Spam, e.g. Good Morning messages.