arXiv subjects

Kun Song

Publications and source records attributed to Kun Song.

At least 19 recordsLinked to original sources

arXiv2025

CollaBot: Vision-Language Guided Simultaneous Collaborative Manipulation

One central goal of robotics is to enable robots to interact with the physical world. Traditional manipulation studies primarily focus on single robots and relatively small objects. However, factory and domestic environments often require large-object manipulation, such as moving tables, where multiple robots must work collaboratively. Existing studies still lack a generalizable framework that can handle diverse objects, tasks, and robot team sizes. In this work, we propose CollaBot, a generalist framework for simultaneous collaborative manipulation. First, we use SEEM for scene segmentation and target-object extraction. Then, we propose a collaborative grasping framework that decomposes the task into local grasp pose generation and global coordination. Finally, we design a two-stage planning module to generate collision-free trajectories for task execution. Experimental results across different settings with varying objects, tasks, and numbers of robots indicate that our framework achieves a 72% success rate. This marks a substantial improvement over behavior cloning-based methods, validating the advantages of the proposed framework in complex multi-robot cooperative tasks. Real-world experiments further demonstrate the feasibility of our method in practical applications.

cs.RO↗

arXiv2025

ProAPO: Progressively Automatic Prompt Optimization for Visual Classification

Vision-language models (VLMs) have made significant progress in image classification by training with large-scale paired image-text data. Their performances largely depend on the prompt quality. While recent methods show that visual descriptions generated by large language models (LLMs) enhance the generalization of VLMs, class-specific prompts may be inaccurate or lack discrimination due to the hallucination in LLMs. In this paper, we aim to find visually discriminative prompts for fine-grained categories with minimal supervision and no human-in-the-loop. An evolution-based algorithm is proposed to progressively optimize language prompts from task-specific templates to class-specific descriptions. Unlike optimizing templates, the search space shows an explosion in class-specific candidate prompts. This increases prompt generation costs, iterative times, and the overfitting problem. To this end, we first introduce several simple yet effective edit-based and evolution-based operations to generate diverse candidate prompts by one-time query of LLMs. Then, two sampling strategies are proposed to find a better initial search point and reduce traversed categories, saving iteration costs. Moreover, we apply a novel fitness score with entropy constraints to mitigate overfitting. In a challenging one-shot image classification setting, our method outperforms existing textual prompt-based methods and improves LLM-generated description methods across 13 datasets. Meanwhile, we demonstrate that our optimal prompts improve adapter-based methods and transfer effectively across different backbones.

cs.CV↗

arXiv2025

Internal Activation Revision: Safeguarding Vision Language Models Without Parameter Update

Vision-language models (VLMs) demonstrate strong multimodal capabilities but have been found to be more susceptible to generating harmful content compared to their backbone large language models (LLMs). Our investigation reveals that the integration of images significantly shifts the model's internal activations during the forward pass, diverging from those triggered by textual input. Moreover, the safety alignments of LLMs embedded within VLMs are not sufficiently robust to handle the activations discrepancies, making the models vulnerable to even the simplest jailbreaking attacks. To address this issue, we propose an \textbf{internal activation revision} approach that efficiently revises activations during generation, steering the model toward safer outputs. Our framework incorporates revisions at both the layer and head levels, offering control over the model's generation at varying levels of granularity. In addition, we explore three strategies for constructing positive and negative samples and two approaches for extracting revision vectors, resulting in different variants of our method. Comprehensive experiments demonstrate that the internal activation revision method significantly improves the safety of widely used VLMs, reducing attack success rates by an average of 48.94\%, 34.34\%, 43.92\%, and 52.98\% on SafeBench, Safe-Unsafe, Unsafe, and MM-SafetyBench, respectively, while minimally impacting model helpfulness.

cs.LG↗

arXiv2024

Enhance Hyperbolic Representation Learning via Second-order Pooling

Hyperbolic representation learning is well known for its ability to capture hierarchical information. However, the distance between samples from different levels of hierarchical classes can be required large. We reveal that the hyperbolic discriminant objective forces the backbone to capture this hierarchical information, which may inevitably increase the Lipschitz constant of the backbone. This can hinder the full utilization of the backbone's generalization ability. To address this issue, we introduce second-order pooling into hyperbolic representation learning, as it naturally increases the distance between samples without compromising the generalization ability of the input features. In this way, the Lipschitz constant of the backbone does not necessarily need to be large. However, current off-the-shelf low-dimensional bilinear pooling methods cannot be directly employed in hyperbolic representation learning because they inevitably reduce the distance expansion capability. To solve this problem, we propose a kernel approximation regularization, which enables the low-dimensional bilinear features to approximate the kernel function well in low-dimensional space. Finally, we conduct extensive experiments on graph-structured datasets to demonstrate the effectiveness of the proposed method.

cs.LG↗

arXiv2024

Exploring Information-Theoretic Metrics Associated with Neural Collapse in Supervised Training

In this paper, we introduce matrix entropy as an analytical tool for studying supervised learning, investigating the information content of data representations and classification head vectors, as well as the dynamic interactions between them during the supervised learning process. Our experimental results reveal that matrix entropy effectively captures the variations in information content of data representations and classification head vectors as neural networks approach Neural Collapse during supervised training, while also serving as a robust metric for measuring similarity among data samples. Leveraging this property, we propose Cross-Model Alignment (CMA) loss to optimize the fine-tuning of pretrained models. To characterize the dynamics of neural networks nearing the Neural Collapse state, we introduce two novel metrics: the Matrix Mutual Information Ratio (MIR) and the Matrix Entropy Difference Ratio (HDR), which quantitatively assess the interactions between data representations and classification heads in supervised learning, with theoretical optimal values derived under the Neural Collapse state. Our experiments demonstrate that MIR and HDR effectively explain various phenomena in neural networks, including the dynamics of standard supervised training, linear mode connectivity. Moreover, we use MIR and HDR to analyze the dynamics of grokking, which is a fascinating phenomenon in supervised learning where a model unexpectedly exhibits generalization long after achieving training data fit.

cs.LG↗

arXiv2024

P2 Explore: Efficient Exploration in Unknown Cluttered Environment with Floor Plan Prediction

Robot exploration aims at the reconstruction of unknown environments, and it is important to achieve it with shorter paths. Traditional methods focus on optimizing the visiting order of frontiers based on current observations, which may lead to local-minimal results. Recently, by predicting the structure of the unseen environment, the exploration efficiency can be further improved. However, in a cluttered environment, due to the randomness of obstacles, the ability to predict is weak. Moreover, this inaccuracy will lead to limited improvement in exploration. Therefore, we propose FPUNet which can be efficient in predicting the layout of noisy indoor environments. Then, we extract the segmentation of rooms and construct their topological connectivity based on the predicted map. The visiting order of these predicted rooms is optimized which can provide high-level guidance for exploration. The FPUNet is compared with other network architectures which demonstrates it is the SOTA method for this task. Extensive experiments in simulations show that our method can shorten the path length by 2.18% to 34.60% compared to the baselines.

cs.RO↗

arXiv2024

Restoring Real-World Degraded Events Improves Deblurring Quality

Due to its high speed and low latency, DVS is frequently employed in motion deblurring. Ideally, high-quality events would adeptly capture intricate motion information. However, real-world events are generally degraded, thereby introducing significant artifacts into the deblurred results. In response to this challenge, we model the degradation of events and propose RDNet to improve the quality of image deblurring. Specifically, we first analyze the mechanisms underlying degradation and simulate paired events based on that. These paired events are then fed into the first stage of the RDNet for training the restoration model. The events restored in this stage serve as a guide for the second-stage deblurring process. To better assess the deblurring performance of different methods on real-world degraded events, we present a new real-world dataset named DavisMCR. This dataset incorporates events with diverse degradation levels, collected by manipulating environmental brightness and target object contrast. Our experiments are conducted on synthetic datasets (GOPRO), real-world datasets (REBlur), and the proposed dataset (DavisMCR). The results demonstrate that RDNet outperforms classical event denoising methods in event restoration. Furthermore, RDNet exhibits better performance in deblurring tasks compared to state-of-the-art methods. DavisMCR are available at https://github.com/Yeeesir/DVS_RDNet.

cs.CV↗

arXiv2024

Distributed Motion Control of Multiple Mobile Manipulators for Reducing Interaction Wrench in Object Manipulation

In real-world cooperative manipulation of objects, multiple mobile manipulator systems may suffer from disturbances and asynchrony, leading to excessive interaction wrenches and potentially causing object damage or emergency stops. Existing methods often rely on torque control and dynamic models, which are uncommon in many industrial robots and settings. Additionally, dynamic models often neglect joint friction forces and are not accurate. These methods are challenging to implement and validate in physical systems. To address the problems, this paper presents a novel distributed motion control approach aimed at reducing these unnecessary interaction wrenches. The control law is only based on local information and joint velocity control to enhance practical applicability. The communication delays within the distributed architecture are considered. The stability of the control law is rigorously proven by the Lyapunov theorem. In the simulations, the effectiveness is shown, and the impact of communication graph connectivity and communication delays has been studied. A comparison with other methods shows the advantages of the proposed control law in terms of convergence speed and robustness. Finally, the control law has been validated in physical experiments. It does not require dynamic modeling or torque control, and thus is more user-friendly for physical robots.

cs.RO↗

arXiv2024

WenetSpeech4TTS: A 12,800-hour Mandarin TTS Corpus for Large Speech Generation Model Benchmark

With the development of large text-to-speech (TTS) models and scale-up of the training data, state-of-the-art TTS systems have achieved impressive performance. In this paper, we present WenetSpeech4TTS, a multi-domain Mandarin corpus derived from the open-sourced WenetSpeech dataset. Tailored for the text-to-speech tasks, we refined WenetSpeech by adjusting segment boundaries, enhancing the audio quality, and eliminating speaker mixing within each segment. Following a more accurate transcription process and quality-based data filtering process, the obtained WenetSpeech4TTS corpus contains $12,800$ hours of paired audio-text data. Furthermore, we have created subsets of varying sizes, categorized by segment quality scores to allow for TTS model training and fine-tuning. VALL-E and NaturalSpeech 2 systems are trained and fine-tuned on these subsets to validate the usability of WenetSpeech4TTS, establishing baselines on benchmark for fair comparison of TTS systems. The corpus and corresponding benchmarks are publicly available on huggingface.

eess.AS↗

arXiv2024

Unveiling the Dynamics of Information Interplay in Supervised Learning

In this paper, we use matrix information theory as an analytical tool to analyze the dynamics of the information interplay between data representations and classification head vectors in the supervised learning process. Specifically, inspired by the theory of Neural Collapse, we introduce matrix mutual information ratio (MIR) and matrix entropy difference ratio (HDR) to assess the interactions of data representation and class classification heads in supervised learning, and we determine the theoretical optimal values for MIR and HDR when Neural Collapse happens. Our experiments show that MIR and HDR can effectively explain many phenomena occurring in neural networks, for example, the standard supervised training dynamics, linear mode connectivity, and the performance of label smoothing and pruning. Additionally, we use MIR and HDR to gain insights into the dynamics of grokking, which is an intriguing phenomenon observed in supervised training, where the model demonstrates generalization capabilities long after it has learned to fit the training data. Furthermore, we introduce MIR and HDR as loss terms in supervised and semi-supervised learning to optimize the information interactions among samples and classification heads. The empirical results provide evidence of the method's effectiveness, demonstrating that the utilization of MIR and HDR not only aids in comprehending the dynamics throughout the training process but can also enhances the training procedure itself.

cs.LG↗

arXiv2024

RHAML: Rendezvous-based Hierarchical Architecture for Mutual Localization

Mutual localization serves as the foundation for collaborative perception and task assignment in multi-robot systems. Effectively utilizing limited onboard sensors for mutual localization between marker-less robots is a worthwhile goal. However, due to inadequate consideration of large scale variations of the observed robot and localization refinement, previous work has shown limited accuracy when robots are equipped only with RGB cameras. To enhance the precision of localization, this paper proposes a novel rendezvous-based hierarchical architecture for mutual localization (RHAML). Firstly, to learn multi-scale robot features, anisotropic convolutions are introduced into the network, yielding initial localization results. Then, the iterative refinement module with rendering is employed to adjust the observed robot poses. Finally, the pose graph is conducted to globally optimize all localization results, which takes into account multi-frame observations. Therefore, a flexible architecture is provided that allows for the selection of appropriate modules based on requirements. Simulations demonstrate that RHAML effectively addresses the problem of multi-robot mutual localization, achieving translation errors below 2 cm and rotation errors below 0.5 degrees when robots exhibit 5 m of depth variation. Moreover, its practical utility is validated by applying it to map fusion when multi-robots explore unknown environments.

cs.RO↗

arXiv2024

Multi-Robot Rendezvous in Unknown Environment with Limited Communication

Rendezvous aims at gathering all robots at a specific location, which is an important collaborative behavior for multi-robot systems. However, in an unknown environment, it is challenging to achieve rendezvous. Previous researches mainly focus on special scenarios where communication is not allowed and each robot executes a random searching strategy, which is highly time-consuming, especially in large-scale environments. In this work, we focus on rendezvous in unknown environments where communication is available. We divide this task into two steps: rendezvous based environment exploration with relative pose (RP) estimation and rendezvous point selection. A new strategy called partitioned and incomplete exploration for rendezvous (PIER) is proposed to efficiently explore the unknown environment, where lightweight topological maps are constructed and shared among robots for RP estimation with very few communications. Then, a rendezvous point selection algorithm based on the merged topological map is proposed for efficient rendezvous for multi-robot systems. The effectiveness of the proposed methods is validated in both simulations and real-world experiments.

cs.RO↗

arXiv2023

A Novel Planning Framework for Complex Flipping Manipulation of Multiple Mobile Manipulators

During complex object manipulation, manipulator systems often face the configuration disconnectivity problem due to closed-chain constraints. Although regrasping can be adopted to get a piecewise connected manipulation, it is a challenging problem to determine whether there is a planning result without regrasping. To address this problem, a novel planning framework is proposed for multiple mobile manipulator systems. Coordinated platform motions and regrasping motions are proposed to enhance configuration connectivity. Given the object trajectory and the grasping pose set, the planning framework includes three steps. First, inverse kinematics for each mobile manipulator is verified along the given trajectory based on different grasping poses. Coverable trajectory segments are determined for each robot for a specific grasping pose. Second, the trajectory choice problem is formulated into a set cover problem, by which we can quickly determine whether the manipulation can be completed without regrasping or with the minimal regrasping number. Finally, the motions of each mobile manipulator are planned with the assigned trajectory segments using existing methods. Both simulations and experimental results show the performance of the planner in complex flipping manipulation. Additionally, the proposed planner can greatly extend the adaptability of multiple mobile manipulator systems in complex manipulation tasks.

cs.RO↗

arXiv2023

FD-Align: Feature Discrimination Alignment for Fine-tuning Pre-Trained Models in Few-Shot Learning

Due to the limited availability of data, existing few-shot learning methods trained from scratch fail to achieve satisfactory performance. In contrast, large-scale pre-trained models such as CLIP demonstrate remarkable few-shot and zero-shot capabilities. To enhance the performance of pre-trained models for downstream tasks, fine-tuning the model on downstream data is frequently necessary. However, fine-tuning the pre-trained model leads to a decrease in its generalizability in the presence of distribution shift, while the limited number of samples in few-shot learning makes the model highly susceptible to overfitting. Consequently, existing methods for fine-tuning few-shot learning primarily focus on fine-tuning the model's classification head or introducing additional structure. In this paper, we introduce a fine-tuning approach termed Feature Discrimination Alignment (FD-Align). Our method aims to bolster the model's generalizability by preserving the consistency of spurious features across the fine-tuning process. Extensive experimental results validate the efficacy of our approach for both ID and OOD tasks. Once fine-tuned, the model can seamlessly integrate with existing methods, leading to performance improvements. Our code can be found in https://github.com/skingorz/FD-Align.

cs.CV↗

arXiv2023

FHT-Map: Feature-based Hierarchical Topological Map for Relocalization and Path Planning

Topological maps are favorable for their small storage compared to geometric map. However, they are limited in relocalization and path planning capabilities. To solve this problem, a feature-based hierarchical topological map (FHT-Map) is proposed along with a real-time map construction algorithm for robot exploration. Specifically, the FHT-Map utilizes both RGB cameras and LiDAR information and consists of two types of nodes: main node and support node. Main nodes will store visual information compressed by convolutional neural network and local laser scan data to enhance subsequent relocalization capability. Support nodes retain a minimal amount of data to ensure storage efficiency while facilitating path planning. After map construction with robot exploration, the FHT-Map can be used by other robots for relocalization and path planning. Experiments are conducted in Gazebo simulator, and the results demonstrate that the proposed FHT-Map can effectively improve relocalization and path planning capability compared with other topological maps. Moreover, experiments on hierarchical architecture are implemented to show the necessity of two types of nodes.

cs.RO↗

arXiv2023

The NPU-MSXF Speech-to-Speech Translation System for IWSLT 2023 Speech-to-Speech Translation Task

This paper describes the NPU-MSXF system for the IWSLT 2023 speech-to-speech translation (S2ST) task which aims to translate from English speech of multi-source to Chinese speech. The system is built in a cascaded manner consisting of automatic speech recognition (ASR), machine translation (MT), and text-to-speech (TTS). We make tremendous efforts to handle the challenging multi-source input. Specifically, to improve the robustness to multi-source speech input, we adopt various data augmentation strategies and a ROVER-based score fusion on multiple ASR model outputs. To better handle the noisy ASR transcripts, we introduce a three-stage fine-tuning strategy to improve translation accuracy. Finally, we build a TTS model with high naturalness and sound quality, which leverages a two-stage framework, using network bottleneck features as a robust intermediate representation for speaker timbre and linguistic content disentanglement. Based on the two-stage framework, pre-trained speaker embedding is leveraged as a condition to transfer the speaker timbre in the source English speech to the translated Chinese speech. Experimental results show that our system has high translation accuracy, speech naturalness, sound quality, and speaker similarity. Moreover, it shows good robustness to multi-source data.

cs.SD↗

arXiv2023

StyleS2ST: Zero-shot Style Transfer for Direct Speech-to-speech Translation

Direct speech-to-speech translation (S2ST) has gradually become popular as it has many advantages compared with cascade S2ST. However, current research mainly focuses on the accuracy of semantic translation and ignores the speech style transfer from a source language to a target language. The lack of high-fidelity expressive parallel data makes such style transfer challenging, especially in more practical zero-shot scenarios. To solve this problem, we first build a parallel corpus using a multi-lingual multi-speaker text-to-speech synthesis (TTS) system and then propose the StyleS2ST model with cross-lingual speech style transfer ability based on a style adaptor on a direct S2ST system framework. Enabling continuous style space modeling of an acoustic model through parallel corpus training and non-parallel TTS data augmentation, StyleS2ST captures cross-lingual acoustic feature mapping from the source to the target language. Experiments show that StyleS2ST achieves good style similarity and naturalness in both in-set and out-of-set zero-shot scenarios.

cs.SD↗

arXiv2023

Gestalt-Guided Image Understanding for Few-Shot Learning

Due to the scarcity of available data, deep learning does not perform well on few-shot learning tasks. However, human can quickly learn the feature of a new category from very few samples. Nevertheless, previous work has rarely considered how to mimic human cognitive behavior and apply it to few-shot learning. This paper introduces Gestalt psychology to few-shot learning and proposes Gestalt-Guided Image Understanding, a plug-and-play method called GGIU. Referring to the principle of totality and the law of closure in Gestalt psychology, we design Totality-Guided Image Understanding and Closure-Guided Image Understanding to extract image features. After that, a feature estimation module is used to estimate the accurate features of images. Extensive experiments demonstrate that our method can improve the performance of existing models effectively and flexibly without retraining or fine-tuning. Our code is released on https://github.com/skingorz/GGIU.

cs.CV↗