Publications

FM-SAM: individual tree crown delineation and classification based on Segmentation Anything Model (SAM) and YOLOv10 in UAV imagery for forest monitoring

Published in Computers and Electronics in Agriculture, 2026

Forest monitoring plays a crucial role in the sustainable management and conservation of natural resources, with tree crown delineation being a key task for assessing forest structure and tree health. This study presents a novel framework, FM-SAM, integrating YOLOv10 and the Segment Anything Model (SAM) for precise tree crown segmentation and species identification in UAV imagery. The framework combines YOLOv10’s real-time detection capabilities with SAM’s strong segmentation, significantly improving segmentation and classification accuracy. SAM is also employed as a semi-supervised branch in FM-SAM to expedite the annotation process, offering a solution that reduces the need for extensive manual labeling. Experimental results on the MixedForestDataset, comprising both coniferous and broadleaf tree species, demonstrate that FM-SAM outperforms traditional deep learning frameworks such as DeepLabv3 and YOLO-based models, achieving high accuracy, precision, and recall values. The proposed method excels in complex forest environments, enhancing the effectiveness of tree crown delineation and classification for large-scale forest monitoring applications.

Recommended citation: Que, H., Gao, H., Shan, W., Liu, M., An, J., Deng, F., ... & Mu, L. (2026). FM-SAM: individual tree crown delineation and classification based on Segmentation Anything Model (SAM) and YOLOv10 in UAV imagery for forest monitoring. Computers and Electronics in Agriculture, 240, 111162.
Download Paper

MACE: Mixture-of-Experts Accelerated Coordinate Encoding for Large-Scale Scene Localization and Rendering

Published in ICRA 2026, 2025

Efficient localization and high-quality rendering in large-scale scenes remain a significant challenge due to the computational cost involved. While Scene Coordinate Regression (SCR) methods perform well in small-scale localization, they are limited by the capacity of a single network when extended to large-scale scenes. To address these challenges, we propose the Mixed Expert-based Accelerated Coordinate Encoding method (MACE), which enables efficient localization and high-quality rendering in large-scale scenes. Inspired by the remarkable capabilities of MOE in large model domains, we introduce a gating network to implicitly classify and select subnetworks, ensuring that only a single sub-network is activated during each inference. Furtheremore, we present AuxiliaryLoss-Free Load Balancing (ALF-LB) strategy to enhance the localization accuracy on large-scale scene. Our framework provides a significant reduction in costs while maintaining higher precision, offering an efficient solution for large-scale scene applications. Additional experiments on the Cambridge test set demonstrate that our method achieves high-quality rendering results with merely 10 minutes of training.

Recommended citation: Liu, M., Fan, D., Que, H., Gao, H., Liu, X., Peng, S., ... & Huang, X. (2025). MACE: Mixture-of-Experts Accelerated Coordinate Encoding for Large-Scale Scene Localization and Rendering. arXiv preprint arXiv:2510.14251.
Download Paper

SenseExpo: Lightweight Neural Networks for Efficient Autonomous Exploration and Scene Prediction

Published in IROS 2025 Workshop, 2025

This work presents SenseExpo, a frontier-based exploration framework powered by a lightweight local map predictor that combines GAN training, a Transformer encoder, and Fast Fourier Convolution. Our smallest model (709k parameters) surpasses much larger baselines (U-Net 24.5M, LaMa 51M) on KTH dataset, achieving PSNR 9.026 and SSIM 0.718, and shows strong cross-domain robustness on HouseExpo (FID 161.55). Leveraging predicted free space for goal selection, SenseExpo accelerates exploration, reducing time by 67.9% on KTH dataset and 77.1% on MRPB 1.0 relative to MapEx, while sustaining high coverage and accuracy. Delivered as a plug-andplay ROS node, it is practical for resource-constrained robots and easy to integrate into existing navigation stacks.

Recommended citation: Que, H., Gao, H., Liu, M., Au, H., Yao, H., & Qiao, F. SenseExpo: Lightweight Neural Networks for Efficient Autonomous Exploration and Scene Prediction. In IROS 2025 Workshop on Edge AI for Robotics: Emerging Technologies and Applications.
Download Paper

Wireless Collaborative Inference Acceleration Based on Distillation for Weed Detection and Instance Segmentation

Published in IROS 2025, 2025

This paper presents a wireless collaborative inference framework optimized for deep learning-based weed instance segmentation on resource-limited weeding robots. Traditional Mask R-CNN struggles with detecting small weeds, suffers from low recall rates, and exhibits the checkerboard effect in segmentation results. To address these challenges, we introduce three key improvements: a feature fusion strategy in the backbone network to enhance small object detection, an improved Region Proposal Network (RPN) with Soft-NMS to reduce false positives and missed detections in complex environments, and a refined mask branch incorporating fully connected upsampling to mitigate checkerboard effects. Additionally, knowledge distillation is employed to compress the model, significantly improving inference speed while maintaining segmentation accuracy. To further enhance inference efficiency, we propose a two-stage approach for determining the optimal partition point and develop a resource-aware optimization algorithm that dynamically adjusts to fluctuating network bandwidth and computational constraints. Experimental evaluations confirm that the proposed approach surpasses existing methods and remains stable across varying resource conditions. A real-world implementation of a drone-server system validates the feasibility of the framework, showcasing its potential for robust and scalable weed detection and segmentation in precision agriculture applications.

A Wireless Collaborated Inference Acceleration Framework for Plant Disease Recognition

Published in ICIC 2025, 2025

Plant disease is a critical factor affecting agricultural production. Traditional manual recognition methods face significant drawbacks, including low accuracy, high costs, and inefficiency. Deep learning techniques have demonstrated significant benefits in identifying plant diseases, but they still face challenges such as inference delays and high energy consumption. Deep learning algorithms are difficult to run on resource-limited embedded devices. Offloading these models to cloud servers is confronted with the restriction of communication bandwidth, and all of these factors will influence the inference’s efficiency. We propose a collaborative inference framework for recognizing plant diseases between edge devices and cloud servers to enhance inference speed. The DNN model for plant disease recognition is pruned through deep reinforcement learning to improve the inference speed and reduce energy consumption. Then the optimal split point is determined by a greedy strategy to achieve the best collaborated inference acceleration. Finally, the system for collaborative inference acceleration in plant disease recognition has been implemented using Gradio to facilitate friendly human-machine interaction. Experiments indicate that the proposed collaborative inference framework significantly increases inference speed while maintaining acceptable recognition accuracy, offering a novel solution for rapidly diagnosing and preventing plant diseases.

Recommended citation: Zhu, H., Huang, X., Gao, H., Jiang, M., Que, H., & Mu, L. (2025, July). A Wireless Collaborated Inference Acceleration Framework for Plant Disease Recognition. In International Conference on Intelligent Computing (pp. 331-341). Singapore: Springer Nature Singapore.
Download Paper

Mapping at First Sense: A Lightweight Neural Network-Based Indoor Structures Prediction Method for Robot Autonomous Exploration

Published in IJCNN 2025, 2025

Autonomous exploration in unknown environments is a critical challenge in robotics, particularly for applications such as indoor navigation, search and rescue, and service robotics. Traditional exploration strategies, such as frontier-based methods, often struggle to efficiently utilize prior knowledge of structural regularities in indoor spaces. To address this limitation, we propose Mapping at First Sense, a lightweight neural network-based approach that predicts unobserved areas in local maps, thereby enhancing exploration efficiency. The core of our method, SenseMapNet, integrates convolutional and transformerbased architectures to infer occluded regions while maintaining computational efficiency for real-time deployment on resourceconstrained robots. Additionally, we introduce SenseMapDataset, a curated dataset constructed from KTH and HouseExpo environments, which facilitates training and evaluation of neural models for indoor exploration. Experimental results demonstrate that SenseMapNet achieves an SSIM (structural similarity) of 0.78, LPIPS (perceptual quality) of 0.68, and an FID (feature distribution alignment) of 239.79, outperforming conventional methods in map reconstruction quality. Compared to traditional frontier-based exploration, our method reduces exploration time by 46.5% (from 2335.56s to 1248.68s) while maintaining a high coverage rate (88%) and achieving a reconstruction accuracy of 88%. The proposed method represents a promising step toward efficient, learning-driven robotic exploration in structured environments.

Recommended citation: Gao, H., Que, H., Li, K., Shan, W., Liu, M., Zhao, R., ... & Qiao, F. (2025). Mapping at First Sense: A Lightweight Neural Network-Based Indoor Structures Prediction Method for Robot Autonomous Exploration. arXiv preprint arXiv:2504.04061.
Download Paper

A Lightweight, Centralized, Collaborative, Truncated Signed Distance Function-Based Dense Simultaneous Localization and Mapping System for Multiple Mobile Vehicles

Published in Sensors, 2024

Simultaneous Localization And Mapping (SLAM) algorithms play a critical role in autonomous exploration tasks requiring mobile robots to autonomously explore and gather information in unknown or hazardous environments where human access may be difficult or dangerous. However, due to the resource-constrained nature of mobile robots, they are hindered from performing long-term and large-scale tasks. In this paper, we propose an efficient multi-robot dense SLAM system that utilizes a centralized structure to alleviate the computational and memory burdens on the agents (i.e. mobile robots). To enable real-time dense mapping of the agent, we design a lightweight and accurate dense mapping method. On the server, to find correct loop closure inliers, we design a novel loop closure detection method based on both visual and dense geometric information. To correct the drifted poses of the agents, we integrate the dense geometric information along with the trajectory information into a multi-robot pose graph optimization problem. Experiments based on pre-recorded datasets have demonstrated our system’s efficiency and accuracy. Real-world online deployment of our system on the mobile vehicles achieved a dense mapping update rate of ∼14 frames per second (fps), a onboard mapping RAM usage of ∼3.4%, and a bandwidth usage of ∼302 KB/s with a Jetson Xavier NX.

Recommended citation: H Que, H Gao, W Shan, X Yang, R Zhao. A Lightweight, Centralized, Collaborative, Truncated Signed Distance Function-Based Dense Simultaneous Localization and Mapping System for Multiple Mobile Vehicles. Sensors. 2024, 24(22):7297. https://doi.org/10.3390/s24227297.
Download Paper