Auto captioning for images is the technique of creating written summaries or captions for images that are relevant to their subject matter. It is a machine learning that blends computer vision (image interpretation) and natural language processing (text production). Automatic image captioning is a relatively new and rapidly increasing academic topic This machine learning role combines natural language processing (text generation) with computer vision (picture interpretation). Although more care is still necessary to attain outcomes comparable to human performance. This work seeks to determine in a systematic manner what distinct and contemporary methodologies and deep neural network models are being utilized for envision captioning. Which techniques are applied to these models? and which tactics have a higher chance of yielding positive results. We conducted a comprehensive literature analysis on current research from utilizing popular databases (Scopus, Web of Sciences, and IEEEXplore). In relation to the goal of this study, we found 61 main studies. We found that language generation is accomplished through RNN or LSTM, whereas CNN is used to analyze visual information and recognize objects in an image. The most often used datasets are flicker 8k, flicker 30k, and MS COCO (used in every session). All study uses the most widely used assessment matrix, BLEU. Additionally it is demonstrated that LSTM with CNN performed better than RNN with CNN. We determined that the concentration mechanism and encoder decoder are the two most promising approaches to implementing this framework, and that combining it can be advantageous. For academics interested in contributing to auto image captioning, this paper offers guidelines and recommendations.