Image Metadata Analysis and Categorization with AI : Enhancing Image Classification Through Metadata Analysis and Deep Learning

In the current digital age, where many images are created and exchanged daily, efficient image classification has become a challenge. Traditional methods have mostly relied on visual content, which proved insufficient. This paper aims to alleviate this issue by developing a system utilizing metadata extraction and machine learning for better classification of images. The first step is to extract metadata from images using ExifTool and prioritize essential fields such as Captions, Sources, and Cities. It then cleans and processes the metadata for accuracy and context. The classification process is further boosted using Term Frequency-Inverse Document Frequency (TF-IDF) vectorization to transform textual information into a machine learning model-informative format. An LSTM neural network is utilized to process the data. The LSTM model was trained with this data and attained 99% accuracy, which indicates how effective it is at processing high-level image information and summarizing and classifying images accurately.