LSDNet: A Lightweight Self-Attentional Distillation Network for Visual Place Recognition

Peng, Guohao; Huang, Yifeng; Li, Heshan; Wu, Zhenyu; Wang, Danwei

doi:10.1109/iros47612.2022.9982272

Cited by 9 publications

(2 citation statements)

References 23 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…2DPass (Yan et al 2022) employs an innovative approach to enhance semantic information extraction from multi-modal data with the integration of two key components: auxiliary modal fusion and multi-scale fusion-to-single distillation. LSD-Net (Peng et al 2022) leverages dual distillation to transfer teacher patterns into students for lightweight VPR. EPC-Net (Hui et al 2022) proposes ProxyConv, which is a lightweight module for local geometric feature aggregation.…”

Section: Knowledge Distillationmentioning

confidence: 99%

DistilVPR: Cross-Modal Knowledge Distillation for Visual Place Recognition

Wang,

She,

Kang

et al. 2024

AAAI

View full text Add to dashboard Cite

The utilization of multi-modal sensor data in visual place recognition (VPR) has demonstrated enhanced performance compared to single-modal counterparts. Nonetheless, integrating additional sensors comes with elevated costs and may not be feasible for systems that demand lightweight operation, thereby impacting the practical deployment of VPR. To address this issue, we resort to knowledge distillation, which empowers single-modal students to learn from cross-modal teachers without introducing additional sensors during inference. Despite the notable advancements achieved by current distillation approaches, the exploration of feature relationships remains an under-explored area. In order to tackle the challenge of cross-modal distillation in VPR, we present DistilVPR, a novel distillation pipeline for VPR. We propose leveraging feature relationships from multiple agents, including self-agents and cross-agents for teacher and student neural networks. Furthermore, we integrate various manifolds, characterized by different space curvatures for exploring feature relationships. This approach enhances the diversity of feature relationships, including Euclidean, spherical, and hyperbolic relationship modules, thereby enhancing the overall representational capacity. The experiments demonstrate that our proposed pipeline achieves state-of-the-art performance compared to other distillation baselines. We also conduct necessary ablation studies to show design effectiveness. The code is released at: https://github.com/sijieaaa/DistilVPR

show abstract

Section: Knowledge Distillationmentioning

confidence: 99%

DistilVPR: Cross-Modal Knowledge Distillation for Visual Place Recognition

Wang,

She,

Kang

et al. 2024

AAAI

View full text Add to dashboard Cite

show abstract

“…This module is necessary when a robot enters a previously mapped environment, as well as when the robot gets lost in the course of moving. Current studies on relocalization can be categorized into image-based [17][18][19][20][21] and lidar-based methods [22][23][24][25][26][27][28].…”

Section: Relocalizationmentioning

confidence: 99%

InfoLa-SLAM: Efficient Lidar-Based Lightweight Simultaneous Localization and Mapping with Information-Based Keyframe Selection and Landmarks Assisted Relocalization

Lin,

Dong,

et al. 2023

Remote Sensing

View full text Add to dashboard Cite

This work reports an information-based landmarks assisted simultaneous localization and mapping (InfoLa-SLAM) in large-scale scenes using single-line lidar. The solution employed two novel designs. The first design was a keyframe selection method based on Fisher information, which reduced the computational cost of the nonlinear optimization for the back-end of SLAM by selecting a relatively small number of keyframes while ensuring the accuracy of mapping. The Fisher information was acquired from the point cloud registration between the current frame and the previous keyframe. The second design was an efficient global descriptor for place recognition, which was achieved by designing a unique graphical feature ID to effectively match the local map with the global one. The results showed that compared with traditional keyframe selection strategies (e.g., based on time, angle, or distance), the proposed method allowed for a 35.16% reduction in the number of keyframes in a warehouse with an area of about 10,000 m2. The relocalization module demonstrated a high probability (96%) of correction even under high levels of measurement noise (0.05 m), while the time consumption for relocalization was below 28 ms. The proposed InfoLa-SLAM was also compared with Cartographer under the same dataset. The results showed that InfoLa-SLAM achieved very similar mapping accuracy to Cartographer but excelled in lightweight performance, achieving a 9.11% reduction in the CPU load and a significant 56.67% decrease in the memory consumption.

show abstract

Fast Dual-Feature Extraction Based on Tightly Coupled Lightweight Network for Visual Place Recognition

Hu,

Zhou,

Lyu

et al. 2023

IEEE Access

View full text Add to dashboard Cite

Visual place recognition (VPR) is a task that aims to predict the location of an image based on the existing images. Because image data can often be massive, extracting features efficiently is critical.To solve the problems of model redundancy and poor time efficiency in feature extraction, this study proposes a fast dual-feature extraction method based on a tightly coupled lightweight network. The tightly coupled network extracts local and global features in a unified model which has a lightweight backbone. Learned step size quantization is then performed to reduce the computational overhead in the inference stage. Additionally, an efficient channel attention module ensures feature representation ability. Efficiency and performance experiments on different hardware platforms showed that the proposed algorithm incurred significant runtime savings for feature extraction, and the inference was 2.9-4.0 times faster than that in the general model. The experimental results confirmed that the proposed method can significantly improve VPR efficiency while ensuring accuracy.

show abstract

LSDNet: A Lightweight Self-Attentional Distillation Network for Visual Place Recognition

Cited by 9 publications

References 23 publications

DistilVPR: Cross-Modal Knowledge Distillation for Visual Place Recognition

DistilVPR: Cross-Modal Knowledge Distillation for Visual Place Recognition

InfoLa-SLAM: Efficient Lidar-Based Lightweight Simultaneous Localization and Mapping with Information-Based Keyframe Selection and Landmarks Assisted Relocalization

Fast Dual-Feature Extraction Based on Tightly Coupled Lightweight Network for Visual Place Recognition

Contact Info

Product

Resources

About