Simple-X a General-Purpose Image Classification Model Through Investigative Approach

In the wider area of artificial intelligence, classification interns to group given item or set of items into their respective categories. This appears to be a simple task overall but as large as your dataset and number of classes grow exponentially it becomes very challenging and exponentially resources heavy process. There are various pre-existed models such as AlexNet, VGG16, VGG19, GoogleNet, MobileNet; these are well proven on various datasets to be efficient. But when we apply them to a specific custom task such as autonomous cars, automation, and robotics, these may not be well suited as they are very time taking and heavy resource demanding like GPU for training on new data. Transfer learning is the most used approach in such cases, which may perform slightly better but not effectively for that specified job. Thus, a need of simple to use and applicable on general-purpose tasks, customizable, and light-weighted neural network architecture is needed. The proposed model is built, by studying the existed models which are been well popular and designing a simple-to-use structure for general-purpose image classification. This proposed neural network can perform moderate with low volumes of data and in primitive resources like time and computing power. This model is very useful to design lightweighted application in edge computing platforms, as this structure is open-sourced and easy to understand and can be modified, added-on, and customized according to your needs.