Image caption generation method based on target detection

Image description is an important task in the field of computer vision and natural language processing, where the goal is to express the information contained in an image in captions. In this paper, we propose an image caption generation model based on the SSD target detection algorithm, which aims to improve caption accuracy. The model first extracts image features using the SSD target detection algorithm and then fuse them with features extracted from the resnet101 network to enhance the accuracy of the image information. Experimental results on the coco dataset show that this SSD target detection algorithm-based model significantly improves the accuracy of the image subtitle generation task and outperforms the traditional model in all five evaluation metrics compared to the traditional model. This research provides a useful reference for further development and practical application in the field of image caption generation.