Deep Learning Based Image Classification Using Small VGG Net Architecture
B. Paulchamy, K. Maharajan, Ramya Srikanteswara · 2022 IEEE 2nd Mysore Sub Section International Conference (MysuruCon) · 2022
One of the most interesting & important areas of computer vision is classification. Deep learning is important in image classification because it allows for the grouping of similar images. Certain adjustments to the images must be made before feeding them to the training model, such as normalization, image resizing, grey-scaling, and so on. Following image pre-processing, the next task is to extract features from the images, draw boxes around the objects in the image, and finally classify the images based on the detection of objects in the images. In a nutshell, classification is a simple task for humans but a difficult task for machines. In this case, we'll use the Small VGGNet (VGG– Visual Geometry Group) Architecture, which is a type of CNN, to classify images using deep learning. The qualified model is validated using three random images from Google. We experimented with scaling the image to different sizes. The deep learning model, which uses Small VGGNet architecture to focus on image classification, performed exceptionally well, correctly categorizing the images. Google Photos has already implemented image classification in their app, but uploading 1000s of images to the cloud requires a strong internet connection because computation is done on Google's server. The issue here is that if we are in a remote location, accessing the internet becomes difficult. Here, we will deploy a trained model with the application to make it easier to classify animals