ZYZecong Ye

Zecong Ye

*This is an algorithmically generated author record based on publicly available information.

Zecong Ye has a total of 35 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    SABDR: Bidirectional Dynamic Domain Adaptation with Style Alignment for Small Object Detection Under Adverse Weather

    by Wei Tang, Xuekai Zhang, Yueping Peng, Hexiang Hao, Zecong Ye, Le Li, Yingying Sun

    Small object detection under adverse weather remains challenging due to weather-induced domain shifts and sparse visual cues of small targets. In contrast to R-YOLO/QTNet and conventional UDA methods, which mainly rely on weather-specific restoration/enhancement or global feature/magnitude alignment, SABDR explicitly targets cross-weather small object adaptation through bidirectional domain translation, degradation-aware receptive-field modeling, feature-statistics modulation, and style-direction alignment. Specifically, the Bidirectional Dynamic Domain Adaptation Network, termed BiDDC-Net, translates between source and target domains and dynamically adjusts receptive fields according to weather severity. The Style-Aware Domain Adaptation Module, termed AIFI-DA, enhances discriminative small-object channels using feature statistics. SDA is further used as a complementary training-time regularizer to encourage style-direction consistency without directly matching feature magnitudes. Experiments are conducted on Cityscapes→Foggy Cityscapes and MOT-Fly→Foggy/Rainy/Snowy MOT-Fly, including newly added rainy and snowy MOT-Fly settings, with both YOLOv5s and YOLO26 evaluated on all MOT-Fly weather conditions. SABDR achieves 47.7 mAP50 on Cityscapes→Foggy Cityscapes, and obtains 96.0%/96.8%, 66.7%/77.1%, and 95.0%/95.6% mAP50 on Foggy, Rainy, and Snowy MOT-Fly with YOLOv5s/YOLO26, respectively. The improvements on MOT-Fly are reported under a fixed single-seed setting and should therefore be interpreted as single-run empirical gains rather than statistically validated improvements. These results demonstrate its effectiveness under the evaluated fog/rain/snow cross-weather small object detection settings.

  • Open Access

    TMRGBT-D2D: A Temporal Misaligned RGB-Thermal Dataset for Drone-to-Drone Target Detection

    by Hexiang Hao, Yueping Peng, Zecong Ye, Baixuan Han, Wei Tang, Wenchao Kang, Xuekai Zhang, Qilong Li, Wenchao Liu

    In the field of drone-to-drone detection tasks, the issue of fusing temporal information with infrared and visible light data for detection has been rarely studied. This paper presents the first temporal misaligned rgb-thermal dataset for drone-to-drone target detection, named TMRGBT-D2D. The dataset covers various lighting conditions (i.e., high-light scenes captured during the day, medium-light and low-light scenes captured at night, with night scenes accounting for 38.8% of all data), different scenes (sky, forests, buildings, construction sites, playgrounds, roads, etc.), different seasons, and different locations, consisting of a total of 42,624 images organized into sequential frames extracted from 19 RGB-T video pairs. Each frame in the dataset has been meticulously annotated, with a total of 94,323 annotations. Except for drones that cannot be identified under extreme conditions, infrared and visible light annotations are one-to-one corresponding. This dataset presents various challenges, including small object detection (the average size of objects in visible light images is approximately 0.02% of the image area), motion blur caused by fast movement, and detection issues arising from imaging differences between different modalities. To our knowledge, this is the first temporal misaligned rgb-thermal dataset for drone-to-drone target detection, providing convenience for research into rgb-thermal image fusion and the development of drone target detection.


  • MBUDet: Misaligned bimodal UAV target detection via target offset label generation

    by Zecong Ye, Hexiang Hao, Yueping Peng, Wei Tang, Xuekai Zhang, Baixuan Han, Haolong Zhai


  • Open Access

    YOLOv11-RGBT: Towards a Comprehensive Single-Stage Multispectral Object Detection Framework

    by Dahang Wan, Rongsheng Lu, Yang Fang, Xianli Lang, Shuangbao Shu, Jingjing Chen, Siyuan Shen, Ting Xu, Zecong Ye

    Multispectral object detection, which integrates information from multiple bands, can enhance detection accuracy and environmental adaptability, holding great application potential across various fields. Although existing methods have made progress in cross-modal interaction, low-light conditions, and model lightweight, there are still challenges like the lack of a unified single-stage framework, difficulty in balancing performance and fusion strategy, and unreasonable modality weight allocation. To address these, based on the YOLOv11 framework, we present YOLOv11-RGBT, a new comprehensive multimodal object detection framework. We designed six multispectral fusion modes and successfully applied them to models from YOLOv3 to YOLOv12 and RT-DETR. After reevaluating the importance of the two modalities, we proposed a P3 mid-fusion strategy and multispectral controllable fine-tuning (MCF) strategy for multispectral models. These improvements optimize feature fusion, reduce redundancy and mismatches, and boost overall model performance. Experiments show our framework excels on three major open-source multispectral object detection datasets, like LLVIP and FLIR. Particularly, the multispectral controllable fine-tuning strategy significantly enhanced model adaptability and robustness. On the FLIR dataset, it consistently improved YOLOv11 models' mAP by 3.41%-5.65%, reaching a maximum of 47.61%, verifying the framework and strategies' effectiveness. The code is available at: https://github.com/wandahangFY/YOLOv11-RGBT.


  • Open Access

    CVNet: Lightweight Cross-View Vehicle ReID with Multi-Scale Localization

    by Wenji Yin, Baixuan Han, Yueping Peng, Hexiang Hao, Zecong Ye, Yu Shen, Yanjun Cai, Wenchao Kang

    Cross-view vehicle re-identification (ReID) between aerial and ground perspectives is challenging due to limited computational resources on edge devices and significant scale variations. We propose CVNet, a lightweight network with two key modules: the multi-scale localization (MSL) module and the deep–shallow filtrate collaboration (DFC) module. The MSL module employs multi-scale depthwise separable convolutions and a localization attention mechanism to extract multi-scale features and localize salient regions, addressing viewpoint variations. DFC employs a dual-branch design comprising deep and shallow branches, integrating a filtration module optimized via neural architecture search, a collaboration module, and lightweight convolutions. This design effectively captures both unique and shared cross-view features, ensuring efficient and robust feature representation. We also release a new CVPair v1.0 dataset, the first benchmark for cross-view ReID, containing 14,969 images of 894 vehicle identities, offering results of traditional and lightweight methods. CVNet achieves state-of-the-art performance on CVPair v1.0, VehicleID, and VeRi776, advancing cross-view vehicle ReID. The dataset will be released publicly.


  • Open Access

    CVNet: Lightweight Cross-View Vehicle ReID with Multi-Scale Localization

    by Wenji Yin, Baixuan Han, Yueping Peng, Hexiang Hao, Zecong Ye, Yu Shen, Yanjun Cai, Wenchao Kang

    Cross-view vehicle re-identification (ReID) between aerial and ground perspectives is challenging due to limited computational resources on edge devices and significant scale variations. We propose CVNet, a lightweight network with two key modules: the multi-scale localization (MSL) module and the deep–shallow filtrate collaboration (DFC) module. The MSL module employs multi-scale depthwise separable convolutions and a localization attention mechanism to extract multi-scale features and localize salient regions, addressing viewpoint variations. DFC employs a dual-branch design comprising deep and shallow branches, integrating a filtration module optimized via neural architecture search, a collaboration module, and lightweight convolutions. This design effectively captures both unique and shared cross-view features, ensuring efficient and robust feature representation. We also release a new CVPair v1.0 dataset, the first benchmark for cross-view ReID, containing 14,969 images of 894 vehicle identities, offering results of traditional and lightweight methods. CVNet achieves state-of-the-art performance on CVPair v1.0, VehicleID, and VeRi776, advancing cross-view vehicle ReID. Code, dataset, and models will be released publicly.

  • Open Access

    A High Performance Air-to-Air Unmanned Aerial Vehicle Target Detection Model

    by Hexiang Hao, Yueping Peng, Zecong Ye, Baixuan Han, Xuekai Zhang, Wei Tang, Wenchao Kang, Qilong Li

    In the air-to-air UAV target detection tasks, the existing algorithms suffer from low precision, low recall and high dependence on device processing power, which makes it difficult to detect UAV small targets efficiently. To solve the above problems, this paper proposes an high-precision model, ATA-YOLOv8. In this paper, we analyze the problem of UAV small target detection from the perspective of the efficient receptive field. The proposed model is evaluated using two air-to-air UAV image datasets, MOT-FLY and Det-Fly, and compared with YOLOv8n and other SOTA algorithms. The experimental results show that the mAP50 of ATA-YOLOv8 is 94.9% and 96.4% on the MOT-FLY and Det-Fly datasets, respectively, which are 25% and 5.9% higher than the mAP of YOLOv8n, while maintaining a model size of 5.1 MB. The methods in this paper improve the accuracy of UAV target detection in air-to-air scenarios. The proposed model’s small size, fast speed and high accuracy make it possible for real-time air-to-air UAV detection on edge-computing devices.


  • Open Access

    Cross-View Multi-Scale Re-Identification Network in the Perspective of Ground Rotorcraft Unmanned Aerial Vehicle

    by Wenji Yin, Yueping Peng, Hexiang Hao, Baixuan Han, Zecong Ye, Wenchao Liu

    Traditional Re-Identification (Re-ID) schemes often rely on multiple cameras from the same perspective to search for targets. However, the collaboration between fixed cameras and unmanned aerial vehicles (UAVs) is gradually becoming a new trend in the surveillance field. Facing the significant perspective differences between fixed cameras and UAV cameras, the task of Re-ID is facing unprecedented challenges. In the setting of a single perspective, although significant advancements have been made in person Re-ID models, their performance markedly deteriorates when confronted with drastic viewpoint changes, such as transitions from aerial to ground-level perspectives. This degradation in performance is primarily attributed to the stark variations between viewpoints and the significant differences in subject posture and background across various perspectives. Existing methods focusing on learning local features have proven to be suboptimal in cross-perspective Re-ID tasks. The reason lies in the perspective distortion caused by the top-down viewpoint of drones, and the richer and more detailed texture information observed from a ground-level perspective, which leads to notable discrepancies in local features. To address this issue, the present study introduces a Multi-scale Across View Model (MAVM) that extracts features at various scales to generate a richer and more robust feature representation. Furthermore, we incorporate a Cross-View Alignment Module (AVAM) that fine-tunes the attention weights, optimizing the model’s response to critical areas such as the silhouette, attire textures, and other key features. This enhancement ensures high recognition accuracy even when subjects change posture and lighting conditions. Extensive experiments conducted on the public dataset AG-ReID have demonstrated the superiority of our proposed method, which significantly outperforms existing state-of-the-art techniques.

  • Open Access

    Research on Low-Slow-Small Unmanned Aerial Vehicle Detection Systems

    by Zecong Ye, Yueping Peng, Hexiang Hao, Baixuan Han, Wenjin Yin, Wenchao Liu


  • Open Access

    An Efficient Adjacent Frame Fusion Mechanism for Airborne Visual Object Detection

    by Zecong Ye, Yueping Peng, Wenchao Liu, Wenji Yin, Hexiang Hao, Baixuan Han, Yanfei Zhu, Dong Xiao

    With the continuous advancement of drone technology, drones are demonstrating a trend toward autonomy and clustering. The detection of airborne objects from the perspective of drones is critical for addressing threats posed by aerial targets and ensuring the safety of drones in the flight process. Despite the rapid advancements in general object detection technology in recent years, the task of object detection from the unique perspective of drones remains a formidable challenge. In order to tackle this issue, our research presents a novel and efficient mechanism for adjacent frame fusion to enhance the performance of visual object detection in airborne scenarios. The proposed mechanism primarily consists of two modules: a feature alignment fusion module and a background subtraction module. The feature alignment fusion module aims to fuse features from aligned adjacent frames and key frames based on their similarity weights. The background subtraction module is designed to compute the difference between the foreground features extracted from the key frame and the background features obtained from the adjacent frames. This process enables a more effective enhancement of the target features. Given that this method can significantly enhance performance without a substantial increase in parameters and computational complexity, by effectively leveraging the feature information from adjacent frames, we refer to it as an efficient adjacent frame fusion mechanism. Experiments conducted on two challenging datasets demonstrate that the proposed method achieves superior performance compared to existing algorithms.