Zining Zhu

Zining Zhu

Auto-generated profile, based on publicly available information

Zining Zhu

Aerospace Information Research Institute, Chinese Academy of Sciences, Beijing, China

Research Focus & Metrics

Academic Career Length

4Years (2023-Present)

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

Representative Publications

View All Publications
4 December 2024 in IEEE Transactions on Geoscience and Remote Sensing
Verified
Journal Article
Closed access
PWPeijin Wang
Peijin Wang
HHHuiyang Hu
Huiyang Hu
BTBoyuan Tong
Boyuan Tong
ZZZiqi Zhang
Ziqi Zhang
FYFanglong Yao
Fanglong Yao
XSXian Sun
Xian Sun
Recently, multi-modal large language models (MLLMs) have shown excellent reasoning capabilities in various fields. Most of the existing remote sensing MLLMs solve image-level text generation problems (e.g., image captioning), but ignore the core issues of object-level recognition, location and multi-temporal changes in the field of remote sensing. In this paper, we propose RingMoGPT, a multi-modal foundation model that unifies vision, language, and localization. Based on the idea of domain adaption, RingMoGPT can complete training by fine-tuning only a few parameters. To make the model capable of object detection and change captioning, we further propose a Location- and Instruction-aware Q-Former and a change detection module, respectively. To improve the performance of RingMoGPT, we carefully design the pre-training dataset and the instruction-tuning dataset. The pre-training dataset contains over a half million high-quality image and text pairs, which are generated through a low-cost and efficient data generation paradigm. The instruction-tuning dataset contains more than 1.6 million question-answer pairs, including 6 downstream tasks: scene classification, object detection, visual question answering, image captioning, grounded image captioning, and change captioning. Our experiments show that RingMoGPT performs well on 6 tasks, especially its ability to analyze multi-temporal data changes and identify dense objects. We also verified the model under zero-shot setting, and the results show that the proposed RingMoGPT also has good generalization ability in the face of new data.

10 January 2023 in ISPRS Journal of Photogrammetry and Remote Sensing
Verified
Journal Article
Closed access
ZZZining Zhu
Zining Zhu
PWPeijin Wang
Peijin Wang
WDWenhui Diao
Wenhui Diao
JYJinze Yang
Jinze Yang
HWHongqi Wang
Hongqi Wang
XSXian Sun
Xian Sun

9 December 2024 in IEEE Transactions on Geoscience and Remote Sensing
Verified
Journal Article
Closed access
ZZZining Zhu
Zining Zhu
PWPeijin Wang
Peijin Wang
WDWenhui Diao
Wenhui Diao
JYJinze Yang
Jinze Yang
LKLingyu Kong
Lingyu Kong
XSXian Sun
Xian Sun
Few-shot object detection (FSOD) has garnered widespread attention in recent years, which makes it possible to learn novel classes with only a handful of labeled samples. Due to the obvious long-tail distribution of remote sensing (RS) data and serious challenges in data labeling, FSOD holds greater practical application value in RS. At present, the FSOD algorithms with fine-tuning of some parameters have attracted much attention due to their stronger incremental learning capacity. However, given large intra-class scale variations and small inter-class feature differences of RS objects, we still need to place greater emphasis on the localization and classification of objects. In this article, we introduce the RoI Feature Refinement (RIFR) method for FSOD in RS imagery, which adopts a powerful training pipeline to better balance attention to the performance of base and novel classes. Aiming at large intra-class scale variations of RS objects, we design a Scale-Aware Feature Compensation Module (SAFCM). By compensating for insufficient scale information, the model’s ability to perceive scale variations of the same class objects has been enhanced. Considering the high similarity among RS classes, we come up with a Prototype Trihard (PT) loss. It achieves the effect of inter-class separability by constraining the relationship between samples and prototypes. Thus, the issue of inter-class confusion has been resultfully solved. Comprehensive experiments on three datasets, DIOR, NWPU VHR-10.v2, and FAIR1M-Airplane, can showcase the efficacy of our RIFR method, and it can implement the most outstanding performance currently. The code will be available at https://github.com/ningerhhh/RIFR.

8 December 2024
Verified
Book Chapter
Closed access
PDF file
JYJinze Yang
Jinze Yang
HWHaoran Wang
Haoran Wang
ZZZining Zhu
Zining Zhu
CLChenglong Liu
Chenglong Liu
MWMeng Wu
Meng Wu
MSMingming Sun
Mingming Sun

17 March 2025 in IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
Verified
Journal Article
Open access
JYJinze Yang
Jinze Yang
Youming WuYouming Wu
Youming Wu
ZZZining Zhu
Zining Zhu
WZWenchao Zhao
Wenchao Zhao
WDWenhui Diao
Wenhui Diao
KFKun Fu
Kun Fu
In the remote sensing field, resolution-degraded (RD) images are often captured under poor imaging conditions, leading to missing texture information for segmentation. To tackle these challenges, a parallel super-resolution branch can be incorporated into the segmentation pipeline to restore detailed features through a feature supplementation operation. However, existing methods based on the structure are mainly limited by the direct supplementation operation between the two tasks with different feature distributions, and the original semantic distribution may be damaged. Therefore, a novel learning method called Generalized Feature Affinity (GFA) is proposed. It is expected to realize more comprehensive feature supplementation in a consistent feature distribution space between the two tasks. Specifically, a Global Distribution Affinity (GDA) module is developed to encourage uniformity in the intra-class distribution between restored and semantic features. Then a Local Texture Affinity (LTA) module is designed to better transfer detailed information by enhancing the texture prominence in feature maps and exploring supplementation in spatial and channel aspects. To guide the supplementation direction more effectively, a Results-oriented Hard Sampling (ROHS) strategy which integrates auxiliary spatial attention is further proposed to enhance the performance of the aforementioned modules. Extensive experiments are conducted on the widely recognized benchmarks ISPRS Potsdam and Vaihingen, and LoveDA datasets. The proposed method results in an improvement of about 2% mIoU across many commonly used segmentation models, demonstrating the effectiveness and generalization of GFA.