Author name cluster

Jinsong Chen

Possible papers associated with this exact author name in Arrow. This page groups case-insensitive exact name matches and is not a full identity disambiguation profile.

4 papers

1 author row

EAAI Journal 2025 Journal Article

Adaptive disentangled target representation for unsupervised domain adaptation in remote sensing segmentation

Runuo Lu
Shoubin Dong
Jianxin Jia
Xusheng Wang
Kai Liu
Jinsong Chen
Shanxin Guo
Xiaorou Zheng

Due to the significant differences between the source and target domains, semantic segmentation models for remote sensing images trained on the source domain often struggle to generalize effectively to new target domains, leading to performance degradation. Most solutions disentangle features into domain-invariant features shared by both domains and domain-specific features unique to each domain, relying primarily on the domain-invariant features for model construction and inference. However, during this process, critical domain-specific features as essential information are often discarded as noise, leading to a reduction in segmentation accuracy. To address these issues, this study proposes an unsupervised domain adaptation, called ChameleonRS 1 1 The code will be publicly available at https: //github. com/SCUT-CCNL/ChameleonRS. . ChameleonRS separates the domain-invariant features and the domain-specific features by disentanglement learning, and introduces a multi-level feature alignment module to enhance the representation of domain-invariant features. More importantly, it aligns predictions based on domain-invariant features with those based on fused invariant and specific features using contrastive learning, effectively capturing target-specific features. This study explores two cross-domain semantic segmentation tasks: (I) Geographic location variation. (II) Geographic location and imaging mode variation. Results show that under setting (I), ChameleonRS outperforms the state-of-the-art method by 4. 79% in Intersection over Union (IoU) and 6. 51% in the F1 score. Under setting (II), it achieves improvements of 4. 46% and 4. 82% in IoU and F1 score. Results demonstrate the superiority and adaptability of ChameleonRS under varying geographical and imaging conditions, effectively mitigating the generalization issues caused by distribution differences, and thereby promoting the development and application of remote sensing technologies in practical scenarios.

Details DOI

NeurIPS Conference 2025 Conference Paper

Rethinking Tokenized Graph Transformers for Node Classification

Jinsong Chen
Chenyang Li
Gaichao Li
John Hopcroft
Kun He

Node tokenized graph Transformers (GTs) have shown promising performance in node classification. The generation of token sequences is the key module in existing tokenized GTs which transforms the input graph into token sequences, facilitating the node representation learning via Transformer. In this paper, we observe that the generations of token sequences in existing GTs only focus on the first-order neighbors on the constructed similarity graphs, which leads to the limited usage of nodes to generate diverse token sequences, further restricting the potential of tokenized GTs for node classification. To this end, we propose a new method termed SwapGT. SwapGT first introduces a novel token swapping operation based on the characteristics of token sequences that fully leverages the semantic relevance of nodes to generate more informative token sequences. Then, SwapGT leverages a Transformer-based backbone to learn node representations from the generated token sequences. Moreover, SwapGT develops a center alignment loss to constrain the representation learning from multiple token sequences, further enhancing the model performance. Extensive empirical results on various datasets showcase the superiority of SwapGT for node classification. Code is available at https: //github. com/JHL-HUST/SwapGT.

PDF Details

NeurIPS Conference 2024 Conference Paper

Leveraging Contrastive Learning for Enhanced Node Representations in Tokenized Graph Transformers

Jinsong Chen
Hanpeng Liu
John E. Hopcroft
Kun He

While tokenized graph Transformers have demonstrated strong performance in node classification tasks, their reliance on a limited subset of nodes with high similarity scores for constructing token sequences overlooks valuable information from other nodes, hindering their ability to fully harness graph information for learning optimal node representations. To address this limitation, we propose a novel graph Transformer called GCFormer. Unlike previous approaches, GCFormer develops a hybrid token generator to create two types of token sequences, positive and negative, to capture diverse graph information. And a tailored Transformer-based backbone is adopted to learn meaningful node representations from these generated token sequences. Additionally, GCFormer introduces contrastive learning to extract valuable information from both positive and negative token sequences, enhancing the quality of learned node representations. Extensive experimental results across various datasets, including homophily and heterophily graphs, demonstrate the superiority of GCFormer in node classification, when compared to representative graph neural networks (GNNs) and graph Transformers.

PDF Details DOI

AAAI Conference 2022 Conference Paper

Towards High-Fidelity Face Self-Occlusion Recovery via Multi-View Residual-Based GAN Inversion

Jinsong Chen
Hu Han
Shiguang Shan

Face self-occlusions are inevitable due to the 3D nature of the human face and the loss of information in the projection process from 3D to 2D images. While recovering face selfocclusions based on 3D face reconstruction, e. g. , 3D Morphable Model (3DMM) and its variants provides an effective solution, most of the existing methods show apparent limitations in expressing high-fidelity, natural, and diverse facial details. To overcome these limitations, we propose in this paper a new generative adversarial network (MvInvert) for natural face self-occlusion recovery without using paired imagetexture data. We design a coarse-to-fine generator for photorealistic texture generation. A coarse texture is computed by inpainting the invisible areas in the photorealistic but incomplete texture sampled directly from the 2D image using the unrealistic but complete statistical texture from 3DMM. Then, we design a multi-view Residual-based GAN Inversion, which re-renders and refines multi-view 2D images, which are used for extracting multiple high-fidelity textures. Finally, these high-fidelity textures are fused based on their visibility maps via Poisson blending. To perform adversarial learning to assure the quality of the recovered texture, we design a discriminator consisting of two heads, i. e. , one for global and local discrimination between the recovered texture and a small set of real textures in UV space, and the other for discrimination between the input image and the rerendered 2D face images via pixel-wise, identity, and adversarial losses. Extensive experiments demonstrate that our approach outperforms the state-of-the-art methods in face selfocclusion recovery under unconstrained scenarios.

PDF Details