Synapse
⌘+K
Synapse
PulseExploreClubsResearchersJournals
Instagram
HomeClubsExplore
May 17, 2026Journal of the Korean Society of Surveying Geodesy Photogrammetry and Cartography

Fusion of Geometric Information and Vision-Language Models for Object Instance Segmentation in Low-Light Aerial 3D Models

View Full Paper
Ask AI
Bookmark
Share

Authors

JLJun Young LeeJLJae Hyoung LimHYH. YunUniversity of Seoul

Discussion

Loading...

Member takes

Implication

Randomized trial demonstrates improved object detection in low-light aerial environments, suggesting enhanced reliability of the framework.

Key Points

  • The study aims to enhance object detection accuracy in low-light aerial environments by integrating geometric information and vision-language models.
  • Developed an object instance segmentation framework combining geometric information and vision-language models.
  • Utilized OpenSeeD for semantic inference in limited visual conditions.
  • Compared performance with ResNeXt-101 based Cascade Mask R-CNN baseline in low-light conditions.
  • Achieved precision improvement of 24.82 percentage points over the baseline model.
  • Increased recall by 34.5 percentage points compared to ResNeXt-101 based Cascade Mask R-CNN.
  • Significantly improved mean Intersection over Union (mIoU) by 37.73 percentage points.

Cite This Study

Lee et al. (2026) studied this question.

synapsesocial.com/papers/6a095b3f7880e6d24efe0fechttps://doi.org/10.7848/ksgpc.2026.44.2.225
View Full Paper
Ask AI
Bookmark
Share

Also Consider

Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context:

  1. 1Geometric Feature-Based Adaptive Voxel Sampling Technique for LiDAR-Based 3D Object Detection2026
  2. 2A Lightweight YOLOv13-G Framework for High-Precision Building Instance Segmentation in Complex UAV Scenes2026
  3. 3LGD-DeepLabV3+: An Enhanced Framework for Remote Sensing Semantic Segmentation via Multi-Level Feature Fusion and Global Modeling2026
  4. 4Fusion-Based Semantic Segmentation and 3D Reconstruction Using Radar–LiDAR Point Clouds: A Comparative Evaluation of DeepLabv3 and FCN-ResNet Against Traditional Architectures2026
  5. 5Video Instance Segmentation Through Hierarchical Offset Compensation and Temporal Memory Update for UAV Aerial Images2025 · 2 citations