Use of Clustering with CNN to Categorize Images inside different directories on OS

In this burgeoning area of Machine Learning and computer vision (CV), image classification plays an eminent role in various real-world problem statements. Advanced implementation of the technology can be used to minimize human interaction with the computers and perform efficiently to overcome the time constraints of the real-world problem. One such application can be used to classify the images stored inside different directories in the operating system and group them on different folders based on their categories. The proposed work focuses on the implementation of this application to overcome the time constraints while searching the images manually inside different directories by the human-computer interaction. This paper demonstrates the use of Keras and Tensorflow to implement CNN on image-based clustering. The convolution operation is used to extract the features from the images. These features are extracted using edge detection, object extraction, and image segmentation. After the features being extracted, the images are clustered (based on their Euclidean distance from other images depending on their features) using the K-means clustering algorithm. To enhance the result, pre-processing is employed on the images for the reduction of noise and elucidate the blurred pixels of the image.