26 March 2026 in IEEE Transactions on Multimedia
Recent Domain Generalized Semantic Segmentation (DGSS) methods have harnessed vision foundation models (VFMs) to achieve remarkable performance in unseen domains. Two key issues require immediate attention from these methods. First, these methods ignore the learning of contextual information, since VFMs are always pretrained on the image classification task, which means that the frozen backbone of these VFMs struggles to extract contextual information used for dense prediction tasks. Second, these methods focus on enhancing domain-invariant information within the learnable tokens or the features extracted by the frozen backbone (e.g., normalizing their style variances), inevitably losing the effective information. To address these two key issues, this paper proposes a Multi-Granularity Token Learning (MGTL) approach to fine-tune VFMs for DGSS, which uses the visual and text embeddings from the CLIP model for optimizing a set of learnable tokens to learn the domain-invariant information of two granularities, including the domain-invariant contextual and conceptual information. Specifically, first, a contextual relation learning strategy is proposed to assist tokens in capturing and embedding contextual information into VFMs, which is crucial to the semantic segmentation task. Second, a conceptual learning strategy is proposed to acquire instance-level conceptual information for the set of learnable tokens and embed it into VFMs to enhance instance-level domain-invariant information. The proposed MGTL approach is evaluated on multiple DGSS benchmarks, achieving superior performance to recent DGSS methods while maintaining low computational cost in trainable parameters during the fine-tuning stage.
1 January 2026 in IEEE Transactions on Multimedia
14 August 2025 in IEEE MultiMedia
Drosophila pose a significant threat to crops, and early detection and identification of Drosophila are crucial. However, the lack of complex Drosophila datasets makes it challenging to use deep learning models in practical scenarios, and traditional data augmentation methods struggle to ensure the quality and effectiveness of generated images. To address the issue of decreased model performance due to dataset scarcity, we propose a deformable two-stage generative adversarial network (DTS-GAN) approach. By synthesizing Drosophila images to augment the original dataset, the accuracy of the recognition model based on the ResNet-50 backbone is significantly improved. This approach enhances the quality and diversity of synthesized images and solves the difficulty of Drosophila identification in complex backgrounds. Experimental results demonstrate that using the proposed DTS-GAN to generate Drosophila images effectively improves the accuracy of the recognition model with an increase from 80.84% to 86.52% compared to the original precision.
1 January 2026 in IEEE Transactions on Geoscience and Remote Sensing
1 January 2026 in IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing