With the rapid evolution of Web 3.0 landscapes, the massive volume of unstructured data emphasizes for the need for a strategic framework which encompasses semantic and knowledge-based methodologies. The requirement of a robust Web 3.0 compatible Image Recommendation framework is imperative to delve through the vast unstructured multimedia content. This paper proposes such a framework, which encompasses metadata generation and utilizes a robust deep learning model to classify the data. The classifier selects features from query enriched knowledge graphs obtained via the Knowledge Graph Search API of Google and Shannon’s entropy is used for feature selection. The model integrates DBSCAN clustering with Lin Similarity in order to generate the intermediate solution set. Semantics oriented learning and reasoning is achieved through Horn’s Index, set to a differential threshold of 0.12 and CoSimRank, set to an initial threshold limit of 0.60 and later increased to 0.70. The proposed model has yielded a precision of 96.22%, recall of 97.84%, accuracy of 97.03%, F-measure of 97.02% and an FDR of 0.04. When compared to similar models it is found that the WISK model performs better than them in all the aforesaid metrics thereby making it one of the best-in-class frameworks for knowledge centric recommendation precision in contrast to its peers.