Xiaofeng Zhang

Dr. Xiaofeng Zhang


Research Focus & Metrics

Academic Career Length

7Years (2020-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
8 June 2025
Verified
Book Chapter
Closed access
PDF file
JBJiesong Bai
Jiesong Bai
YYYuhao Yin
Yuhao Yin
QHQiyuan He
Qiyuan He
YLYuanxian Li
Yuanxian Li
Xiaofeng ZhangXiaofeng Zhang
Xiaofeng Zhang

26 August 2020 in IEEE Signal Processing Letters
Verified
Journal Article
Closed access
Xiaofeng ZhangXiaofeng Zhang
Xiaofeng Zhang
FCFeng Chen
Feng Chen
CWCailing Wang
Cailing Wang
MTMing Tao
Ming Tao
GJGuo-Ping Jiang
Guo-Ping Jiang
Different from image inpainting, image outpainting has relatively less context in the image center to capture and more content at the image border to predict. Therefore, classical encoder-decoder pipeline of existing methods may not predict the outstretched unknown content perfectly. In this paper, a novel two-stage siamese adversarial model for image extrapolation, named Siamese Expansion Network (SiENet) is proposed. Specifically, in two stages, a novel border sensitive convolution named adaptive filling convolution is designed for allowing encoder to predict the unknown content, alleviating the burden of decoder. Besides, to introduce prior knowledge to network and reinforce the inferring ability of encoder, siamese adversarial mechanism is designed to enable our network to model the distribution of covered long range feature as that of uncovered image feature. The results on four datasets has demonstrated that our method outperforms existing state-of-the-arts and could produce realistic results. Our code is released on https://github.com/nanjingxiaobawang/SieNet-Image-extrapolation

22 December 2022 in Remote Sensing
Verified
Journal Article
Open access
PDF file
XMXinkui Mei
Xinkui Mei
Xiufen YeXiufen Ye
Xiufen Ye
Xiaofeng ZhangXiaofeng Zhang
Xiaofeng Zhang
YLYusong Liu
Yusong Liu
JWJunting Wang
Junting Wang
XWXuli Wang
Xuli Wang
Because of the unique physical and chemical properties of water, obtaining high-quality underwater images directly is not an easy thing. Hence, recovery and enhancement are indispensable steps in underwater image processing and have therefore become research hotspots. Nevertheless, existing image-processing methods generally have high complexity and are difficult to deploy on underwater platforms with limited computing resources. To tackle this issue, this paper proposes a simple and effective baseline named UIR-Net that can recover and enhance underwater images simultaneously. This network uses a channel residual prior to extract the channel of the image to be recovered as a prior, combined with a gradient strategy to reduce parameters and training time to make the operation more lightweight. This method can improve the color performance while maintaining the style and spatial texture of the contents. Through experiments on three datasets (MSRB, MSIRB and UIEBD-Snow), we confirm that UIR-Net can recover clear underwater images from original images with large particle impurities and ocean light spots. Compared to other state-of-the-art methods, UIR-Net can recover underwater images at a similar or higher quality with a significantly lower number of parameters, which is valuable in real-world applications.

31 December 2024 in Neural Networks
Verified
Journal Article
Closed access
Xiaofeng ZhangXiaofeng Zhang
Xiaofeng Zhang
FZFanshuo Zeng
Fanshuo Zeng
CGChaochen Gu
Chaochen Gu

28 October 2024
Verified
Conference Paper
Open access
JWJinfeng Wei
Jinfeng Wei
Xiaofeng ZhangXiaofeng Zhang
Xiaofeng Zhang
In this work, we introduce DOPRA, a novel approach designed to mitigate hallucinations in multi-modal large language models (MLLMs). Unlike existing solutions that typically involve costly supplementary training data or the integration of external knowledge sources, DOPRA innovatively addresses hallucinations by decoding specific weighted layer penalties and redistribution, offering an economical and effective solution without additional resources. DOPRA is grounded in unique insights into the intrinsic mechanisms controlling hallucinations within MLLMs, especially the models' tendency to over-rely on a subset of summary tokens in the self-attention matrix, neglecting critical image-related information. This phenomenon is particularly pronounced in certain strata. To counteract this over-reliance, DOPRA employs a strategy of weighted overlay penalties and redistribution in specific layers, such as the 12th layer, during the decoding process. Furthermore, DOPRA includes a retrospective allocation process that re-examines the sequence of generated tokens, allowing the algorithm to reallocate token selection to better align with the actual image content, thereby reducing the incidence of hallucinatory descriptions in auto-generated captions. Overall, DOPRA represents a significant step forward in improving the output quality of MLLMs by systematically reducing hallucinations through targeted adjustments during the decoding process.

11 April 2025 in Proceedings of the AAAI Conference on Artificial Intelligence
Verified
Journal Article
Open access
PDF file
Xiaofeng ZhangXiaofeng Zhang
Xiaofeng Zhang
FZFanshuo Zeng
Fanshuo Zeng
YQYihao Quan
Yihao Quan
ZHZheng Hui
Zheng Hui
JYJiawei Yao
Jiawei Yao
Multimodal large language models have experienced rapid growth, and numerous different models have emerged. The interpretability of LVLMs remains an under-explored area. Especially when faced with more complex tasks such as chain-of-thought reasoning, its internal mechanisms still resemble a black box that is difficult to decipher. By studying the interaction and information flow between images and text, we noticed that in models such as LLaVA1.5, image tokens that are semantically related to text are more likely to have information flow convergence in the LLM decoding layer, and these image tokens receive higher attention scores. However, those image tokens that are less relevant to the text do not have information flow convergence, and they only get very small attention scores. To efficiently utilize the image information, we propose a new image token reduction method, Simignore, which aims to improve the complex reasoning ability of LVLMs by computing the similarity between image and text embeddings and ignoring image tokens that are irrelevant and unimportant to the text. Through extensive experiments, we demonstrate the effectiveness of our method for complex reasoning tasks.