We present Flat'n'Fold, a novel large-scale dataset for garment manipulation that addresses critical gaps in existing datasets. Comprising 1,212 human and 887 robot demonstrations of flattening and folding 44 unique garments across 8 categories, Flat'n'Fold surpasses prior datasets in size, scope, and diversity. Our dataset uniquely captures the entire manipulation process from crumpled to folded states, providing synchronized multi-view RGB-D images, point clouds, and action data, including hand or gripper positions and rotations. We quantify the dataset's diversity and complexity compared to existing benchmarks and show that our dataset features natural and diverse manipulations of real-world demonstrations of human and robot demonstrations in terms of visual and action information. To showcase Flat'n'Fold's utility, we establish new benchmarks for grasping point prediction and subtask decomposition. Our evaluation of state-of-the-art models on these tasks reveals significant room for improvement. This underscores Flat'n'Fold's potential to drive advances in robotic perception and manipulation of deformable objects. Our dataset can be downloaded at https://cvas-ug.github.io/flat-n-fold

This paper presents a rigorous evaluation of Real-to-Sim parameter estimation approaches for fabric manipulation in robotics. The study systematically assesses three state-of-the-art approaches, namely two differential pipelines and a data-driven approach. We also devise a novel physics-informed neural network approach for physics parameter estimation. These approaches are interfaced with two simulations across multiple Real-to-Sim scenarios (lifting, wind blowing, and stretching) for five different fabric types and evaluated on three unseen scenarios (folding, fling, and shaking). We found that the simulation engines and the choice of Real-to-Sim approaches significantly impact fabric manipulation performance in our evaluation scenarios. Moreover, PINN observes superior performance in quasi-static tasks but shows limitations in dynamic scenarios. Videos and source code are available at cvas-ug.github.io/real2sim-study.

Interactive Text-to-image retrieval (I-TIR) is an important enabler for a wide range of state-of-the-art services in domains such as e-commerce and education.However, current methods rely on finetuned Multimodal Large Language Models (MLLMs), which are costly to train and update, and exhibit poor generalizability.This latter issue is of particular concern, as: 1) finetuning narrows the pretrained distribution of MLLMs, thereby reducing generalizability; and 2) I-TIR introduces increasing query diversity and complexity.As a result, I-TIR solutions are highly likely to encounter queries and images not well represented in any training dataset.To address this, we propose leveraging Diffusion Models (DMs) for text-to-image mapping, to avoid finetuning MLLMs while preserving robust performance on complex queries.Specifically, we introduce Diffusion Augmented Retrieval (DAR), a framework that generates multiple intermediate representations via LLM-based dialogue refinements and DMs, producing a richer depiction of the user's information needs.This augmented representation facilitates more accurate identification of semantically and visually related images.Extensive experiments on four benchmarks show that for simple queries, DAR achieves results on par with finetuned I-TIR models, yet without incurring their tuning overhead.Moreover, as queries become more complex through additional conversational turns, DAR surpasses finetuned I-TIR models by up to 7.61% in Hits@10 after ten turns, illustrating its improved generalization for more intricate queries.

State-of-the-art pre-trained image models predominantly adopt a two-stage approach: initial unsupervised pre-training on large-scale datasets followed by task-specific fine-tuning using Cross-Entropy loss (CE). However, it has been demonstrated that CE can compromise model generalization and stability. While recent works employing contrastive learning address some of these limitations by enhancing the quality of embeddings and producing better decision boundaries, they often overlook the importance of hard negative mining and rely on resource intensive and slow training using large sample batches. To counter these issues, we introduce a novel approach named CLCE, which integrates Label-Aware Contrastive Learning with CE. Our approach not only maintains the strengths of both loss functions but also leverages hard negative mining in a synergistic way to enhance performance. Experimental results demonstrate that CLCE significantly outperforms CE in Top-1 accuracy across twelve benchmarks, achieving gains of up to 3.52% in few-shot learning scenarios and 3.41% in transfer learning settings with the BEiT-3 model. Importantly, our proposed CLCE approach effectively mitigates the dependency of contrastive learning on large batch sizes such as 4096 samples per batch, a limitation that has previously constrained the application of contrastive learning in budget-limited hardware environments.

As Multimodal Large Language Models (MLLMs) grow in size, adapting them to specialized tasks becomes increasingly challenging due to high computational and memory demands. Indeed, traditional fine-tuning methods are costly, due to the need for extensive, task-specific training. While efficient adaptation methods exist that aim to reduce these costs, in practice they suffer from shallow inter-modal alignment, which severely hurts model effectiveness. To tackle these computational challenges and improve inter-modal alignment, we introduce the MultiWay-Adapter (MWA), a novel framework featuring an ‘Alignment Enhancer’. This enhancer deepens inter-modal alignment, enabling high transferability with minimal tuning effort. Our experiments show that unlike prior efficient tuning approaches, MWA maintains model effectiveness, while reducing training time by up-to 57%. MWA is also lightweight, increasing model size by only 2-3% (in terms of parameters) for state-of-the-art foundation models like BEiT-3 Large. These results demonstrate that MWA provides an efficient and effective adaptation method for MLLMs, significantly broadening their applicability.

Vision Transformers (ViTs) have emerged as popular models in computer vision, demonstrating state-of-the-art performance across various tasks. This success typically follows a two-stage strategy involving pre-training on large-scale datasets using self-supervised signals, such as masked random patches, followed by fine-tuning on task-specific labeled datasets with cross-entropy loss. However, this reliance on cross-entropy loss has been identified as a limiting factor in ViTs, affecting their generalization and transferability to downstream tasks. Addressing this critical challenge, we introduce a novel Label-aware Contrastive Training framework, LaCViT, which significantly enhances the quality of embeddings in ViTs. LaCViT not only addresses the limitations of cross-entropy loss but also facilitates more effective transfer learning across diverse image classification tasks. Our comprehensive experiments on eight standard image classification datasets reveal that LaCViT statistically significantly enhances the performance of three evaluated ViTs by up-to 10.78% under Top-1 Accuracy.

Teleoperating robotic arms can be a challenging task for non-experts, particularly when using complex control devices or interfaces. To address the limitations and challenges of existing teleoperation frameworks, such as cognitive strain, control complexity, robot compatibility, and user evaluation, we propose TELESIM, a modular and plug-and-play framework that enables direct teleoperation of any robotic arm using a digital twin as the interface between users and the robotic system. Due to TELESIM’s modular design, it is possible to control the digital twin using any device that outputs a 3D pose, such as a virtual reality controller or a finger-mapping hardware controller. To evaluate the efficacy and user-friendliness of TELESIM, we conducted a user study with 37 participants. The study involved a simple pick-and-place task, which was performed using two different robots equipped with two different control modalities. Our experimental results show that most users were able to succeed by building at least a tower of 3 cubes in 10 minutes, with only 5 minutes of training beforehand, regardless of the control modality or robot used, demonstrating the usability and user-friendliness of TELESIM.

Robotic vision applications often necessitate a wide range of visual perception tasks, such as object detection, segmentation, and identification. While there have been substantial advances in these individual tasks, integrating specialized models into a unified vision pipeline presents significant engineering challenges and costs. Recently, Multimodal Large Language Models (MLLMs) have emerged as novel backbones for various downstream tasks. We argue that leveraging the pre-training capabilities of MLLMs enables the creation of a simplified framework, thus mitigating the need for task-specific encoders. Specifically, the large-scale pretrained knowledge in MLLMs allows for easier fine-tuning to downstream robotic vision tasks and yields superior performance. We introduce the RoboLLM framework, equipped with a BEiT-3 backbone, to address all visual perception tasks in the ARMBench challenge—a large-scale robotic manipulation dataset about real-world warehouse scenarios. RoboLLM not only outperforms existing baselines but also substantially reduces the engineering burden associated with model selection and tuning. All the code used in this paper can be found in https://github.com/longkukuhi/RoboLLM.

Abstract This study addresses the problem of hand–eye calibration in robotic systems by developing Continual Learning (CL)-based approaches. Traditionally, robots require explicit models to transfer knowledge from camera observations to their hands or base. However, this poses limitations, as the hand–eye calibration parameters are typically valid only for the current camera configuration. We, therefore, propose a flexible and autonomous hand–eye calibration system that can adapt to changes in camera pose over time. Three CL-based approaches are introduced: the naive CL approach, the reservoir rehearsal approach, and the hybrid approach combining reservoir sampling with new data evaluation. The naive CL approach suffers from catastrophic forgetting, while the reservoir rehearsal approach mitigates this issue by sampling uniformly from past data. The hybrid approach further enhances performance by incorporating reservoir sampling and assessing new data for novelty. Experiments conducted in simulated and real-world environments demonstrate that the CL-based approaches, except for the naive approach, achieve competitive performance compared to traditional batch learning-based methods. This suggests that treating hand–eye calibration as a time sequence problem enables the extension of the learned space without complete retraining. The adaptability of the CL-based approaches facilitates accommodating changes in camera pose, leading to an improved hand–eye calibration system.

We present a behaviour-based reinforcement learning approach, inspired by Brook's subsumption architecture, in which simple fully connected networks are trained as reactive behaviours. Our working assumption is that a pick and place robotic task can be simplified by leveraging domain knowledge of a robotics developer to decompose and train reactive behaviours; namely, approach, grasp, and retract. Then the robot autonomously learns how to combine reactive behaviours via an Actor-Critic architecture. We use an Actor-Critic policy to determine the activation and inhibition mechanisms of the reactive behaviours in a particular temporal sequence. We validate our approach in a simulated robot environment where the task is about picking a block and taking it to a target position while orienting the gripper from a top grasp. The latter represents an extra degree-of-freedom of which current end-to-end reinforcement learning approaches fail to generalise. Our findings suggest that robotic learning can be more effective if each behaviour is learnt in isolation and then combined them to accomplish the task. That is, our approach learns the pick and place task in 8,000 episodes, which represents a drastic reduction in the number of training episodes required by an end-to-end approach ( 95,000 episodes) and existing state-of-the-art algorithms.

This paper proposes a single-shot approach for recognising clothing categories from 2.5D features. We propose two visual features, BSP (B-Spline Patch) and TSD (Topology Spatial Distances) for this task. The local BSP features are encoded by LLC (Locality-constrained Linear Coding) and fused with three different global features. Our visual feature is robust to deformable shapes and our approach is able to recognise the category of unknown clothing in unconstrained and random configurations. We integrated the category recognition pipeline with a stereo vision system, clothing instance detection, and dual-arm manipulators to achieve an autonomous sorting system. To verify the performance of our proposed method, we build a high-resolution RGBD clothing dataset of 50 clothing items of 5 categories sampled in random configurations (a total of 2,100 clothing samples). Experimental results show that our approach is able to reach 83.2% accuracy while classifying clothing items which were previously unseen during training. This advances beyond the previous state-of-the-art by 36.2%. Finally, we evaluate the proposed approach in an autonomous robot sorting system, in which the robot recognises a clothing item from an unconstrained pile, grasps it, and sorts it into a box according to its category. Our proposed sorting system achieves reasonable sorting success rates with single-shot perception.

In this paper, we propose a Gaussian Process-based interactive perception approach for recognising highly-wrinkled clothes. We have integrated this recognition method within a clothes sorting pipeline for the pre-washing stage of an autonomous laundering process. Our approach differs from reported clothing manipulation approaches by allowing the robot to update its perception confidence via numerous interactions with the garments. The classifiers predominantly reported in clothing perception (e.g. SVM, Random Forest) studies do not provide true classification probabilities, due to their inherent structure. In contrast, probabilistic classifiers (of which the Gaussian Process is a popular example) are able to provide predictive probabilities. In our approach, we employ a multi-class Gaussian Process classification using the Laplace approximation for posterior inference and optimising hyper-parameters via marginal likelihood maximisation. Our experimental results show that our approach is able to recognise unknown garments from highly-occluded and wrinkled configurations and demonstrates a substantial improvement over non-interactive perception approaches.

We present a visually guided, dual-arm, industrial robot system that is capable of autonomously flattening garments by means of a novel visual perception pipeline that fully interprets high-quality RGB-D images of a clothing scene based on an active stereo robot head. A segmented clothing range map is B-Spline smoothed prior to being parsed by means of shape and topology analysis into ‘wrinkle’ structures. The length, width and height of each wrinkle is used to quantify the topology of each wrinkle and thereby rank wrinkles by size such that a greedy algorithm can identify the largest wrinkle present. A flattening plan optimised for the largest detected wrinkle is formulated based on dual-arm manipulation. We report the validation of our autonomous flattening behaviour and observe that dual-arm flattening requires significantly fewer manipulation iterations than single-arm flattening. Our experimental results also reveal that the flattening process is heavily influenced by the quality of the RGB-D sensor: use of a custom off-the-shelf high-resolution stereo-based sensor system outperformed a commercial low-resolution kinect-like camera in terms of required flattening iterations.