The practice of generating a textual explanation for image is called as image captioning. Currently, it is one of the recent and growing research problems. Caption generation is helpful in assistive technologies, content moderation, e-commerce websites apart from social media. In this work, a module has been designed for the cluster of relevant captions based on the given image. For enhancing the relevance of the caption generation mechanism, two models, namely CNN and LSTM, have been used that perform as encoder and decoder, respectively. Further, the results are evaluated using an evaluation metric called BLEU score. For this work, Flickr 8 k online dataset has been used. The results show that in case of input images consists of the single and double objects the model performed considerably well as compared to multiple objects images. The average BLEU score of images having single, double, and multiple objects is 0.61, 0.60, and 0.38, respectively. As a result, the methodology contributes to the development of advanced models capable of producing acceptable caption creation outcomes.