Published Paper Details:

Image and video caption generation has gained significant attention in recent years due to its potential in enhancing content accessibility, search ability, and user experience. Deep learning techniques have shown remarkable success in addressing this task by leveraging the power of convolutional neural networks (CNNs) and recurrent neural networks (RNNs). In this paper, we propose a novel approach for image and video caption generation using deep learning. Our model consists of two key components: an image/video encoder and a caption generator. The image/video encoder utilizes a pre-trained CNN, such as ResNet or VGG, to extract high-level visual features from the input image or video frames. These features are then fed into an RNN-based caption generator, which sequentially generates a caption word-by-word, taking into account both the visual features and the context provided by the previously generated words. To further enhance the caption generation process, we incorporate attention mechanisms that allow the model to focus on relevant regions or frames while generating each word. This attention mechanism helps the model capture fine-grained details and improves the overall quality and coherence of the generated captions. To train our model, we use large-scale image and video caption datasets, such as MSCOCO and MSVD, which provide extensive annotations for training and evaluation. Experimental results demonstrate that our proposed approach achieves state-of-the-art performance on several benchmark datasets, surpassing existing methods in terms of caption quality and diversity.