Fusion: Practice and Applications FPA 2692-4048 2770-0070 10.54216/FPA https://www.americaspg.com/journals/show/730 2018 2018 Bharati Vidyapeeth’s College of Engineering, New Delhi, India Shitiz Gupta Bharati Vidyapeeth’s College of Engineering, New Delhi, India Shubham Agnihotri Bharati Vidyapeeth’s College of Engineering, New Delhi, India Deepasha Birla Bharati Vidyapeeth’s College of Engineering, New Delhi, India Achin Jain College of computer engineering and sciences, Prince Sattam bin abdulaziz University, Saudi Arabia Thavavel Vaiyapuri Bharati Vidyapeeth’s College of Engineering, New Delhi, India Puneet Singh Lamba Image caption generation is a stimulating multimodal task. Substantial advancements have been made in thefield of deep learning notably in computer vision and natural language processing. Yet, human-generated captions are still considered better, which makes it a challenging application for interactive machine learning. In this paper, we aim to compare different transfer learning techniques and develop a novel architecture to improve image captioning accuracy. We compute image feature vectors using different state-of-the-art transferlearning models which are fed into an Encoder-Decoder network based on Stacked LSTMs with soft attention,along with embedded text to generate high accuracy captions. We have compared these models on severalbenchmark datasets based on different evaluation metrics like BLEU and METEOR. 2021 2021 42 55 10.54216/FPA.040202 https://www.americaspg.com/articleinfo/3/show/730