- Conference Article
- 10.1145/3641584.3641588
Text generation image method based on convolutional attention and cross-domain feature fusion
- Sep 22, 2023
- Pingan Qiao + 2 more +2
Text-generated images are one of the tasks of multimodal machine learning. Although the images generated by previous algorithms can meet the requirements of text description, the generated images are not clear enough and the processing of image details still needs to be improved. Therefore, in this paper, we propose CF-AttnGAN (Convolutional attention and cross-domain feature fusion for generative adversarial networks) model based on AttnGAN to improve the quality of generated images and enrich the texture features of the generated images. In the initial image generation stage, the Convolutional Block Attention Module (CBAM) is proposed, which is an effective and simple feedforward convolutional neural network attention mechanism that enables the input features to be refined in both channel and spatial dimensions by convolutional operations to deeply text representation capability of this network is mined. In the refinement stage of the initial image, a Feature Fusion Strengthen Response Module (FFSRM) is proposed to improve the semantic consistency of text features with visual features in order to allow the fine-grained generator to better capture the semantic and image information. Extensive experiments on the CUB-200-2011 dataset show that the CF-AttnGAN model can generate higher quality images.
Read more