In today’s digital era recommendation system is widely used applications from Netflix to Amazon, from Google to Goodreads, etc. Amazon estimated about 35% incremental revenue from product recommendations every year. Majority of e-commerce companies provides a recommendation system for purchasing product. The most search engine uses text-based search criteria for finding similar products. In this article, a comprehensive study of research contents related to text and image-based recommendation system is mentioned. First, we presented various text pre-processing techniques are used to clean dataset. Second, we described various techniques used for finding textual based similarity. Third, an advanced technique likeConvolution Neural Network(ConvNets) is used for computing image-based similarity of product. Cosine similarity is calculated when word occurrence is important in textual input whereas Euclidean distance is measured when an image or semantic-based text is considered as input. Experimental results depict how recommendations using ConvNets is more exciting than usual classical techniques.