What models are best for image classification? A research report.
There is a lot of applications of image classification. In the agricultural sector it’s of interest to identify crops from weeds based on image classification, in the advancement of self driven cars we want to be able to detect and classify images in front of these cars; is it a human? some nylon flying about? or some animal.
This research is done to find out best models for image classification in order to detect hotels. So we want to classify hotel and non-hotel building or rather images.
From the keras application page, available models are compared with details of their Top-1 Accuracy, Top-5 Accuracy, Parameters and Depth.
These models require a large data set to be trained and are computationally expensive. Hence a good approach the pre-trained model. This saves time and solves the problem of getting the large data set and required hardware for solving a specific problem . Since we can pre-train the models with available large data set and use the pre-trained model for the specific problem at hand.