Image Caption Generation using Deep Residual Learning

The process of creating a written description of an image that describes the action depicted in it is known as image captioning. It is one of the most challenging study areas, and the only method that can help is the machine learning one. Image captioning is a tool where intellectual approach is required for better performance. The suggested work uses CNN, RNN, and Deep Residual Network to propose an image captioning system. RNN serves as the decoder while CNN serves as the encoder. For the MSCOCO benchmark, the system can accurately infer the state of affairs. Deep Residual Network (Deep ResNet), which efficiently utilises the layers and cuts down on computation time, was used to train the model. Deep ResNet, also known as skip connection, bypasses the convolutional layers that provided a solution to the gradient explosion problem. Comparing the Bilingual Evaluation Understudy (BLEU) score to the previously used model, the system perceived a higher score. By lowering the image's size and improving its visibility return good results.