Kazuki Kozuka

Xudong Wang, Shufan Li, Konstantinos Kallidromitis et al.

Open-vocabulary image segmentation aims to partition an image into semantic regions according to arbitrary text descriptions. However, complex visual scenes can be naturally decomposed into simpler parts and abstracted at multiple levels of granularity, introducing inherent segmentation ambiguity. Unlike existing methods that typically sidestep this ambiguity and treat it as an external factor, our approach actively incorporates a hierarchical representation encompassing different semantic-levels into the learning process. We propose a decoupled text-image fusion mechanism and representation learning modules for both "things" and "stuff". Additionally, we systematically examine the differences that exist in the textual and visual features between these types of categories. Our resulting model, named HIPIE, tackles HIerarchical, oPen-vocabulary, and unIvErsal segmentation tasks within a unified framework. Benchmarked on over 40 datasets, e.g., ADE20K, COCO, Pascal-VOC Part, RefCOCO/RefCOCOg, ODinW and SeginW, HIPIE achieves the state-of-the-art results at various levels of image comprehension, including semantic-level (e.g., semantic segmentation), instance-level (e.g., panoptic/referring segmentation and object detection), as well as part-level (e.g., part/subpart segmentation) tasks. Our code is released at https://github.com/berkeley-hipie/HIPIE.

Shufan Li, Konstantinos Kallidromitis, Akash Gokul Yusuke Kato et al.

Group-advantage-based reinforcement learning methods, such as GRPO and DAPO, have demonstrated strong performance across diverse domains, including mathematical reasoning and text-to-image generation. However, their reliance on sample-level rewards introduces a key limitation as uniform credit assignment across all tokens fails to capture fine-grained, token-level contributions. To address this issue, we propose Guidance Contrastive Policy Optimization (GCPO), a novel algorithm that enables per-token credit assignment by contrasting model predictions under positive and negative prompts. Rather than uniformly broadcasting sample-level advantages, GCPO assigns token-level advantages proportional to the difference between these contrastive predictions, allowing more precise and informative learning signals. Empirically, we find that GCPO emphasizes semantically relevant regions such as visual areas aligned with textual prompts in text-to-image generation, and critical keywords within reasoning traces for chain-of-thought tasks. Through extensive experiments, GCPO consistently outperforms GRPO and DAPO baselines on both text-to-image generation and chain-of-thought reasoning benchmarks, demonstrating its effectiveness as a general and scalable optimization strategy for discrete policy learning.

Akash Gokul, Konstantinos Kallidromitis, Shufan Li et al.

Recent works in self-supervised learning have demonstrated strong performance on scene-level dense prediction tasks by pretraining with object-centric or region-based correspondence objectives. In this paper, we present Region-to-Object Representation Learning (R2O) which unifies region-based and object-centric pretraining. R2O operates by training an encoder to dynamically refine region-based segments into object-centric masks and then jointly learns representations of the contents within the mask. R2O uses a "region refinement module" to group small image regions, generated using a region-level prior, into larger regions which tend to correspond to objects by clustering region-level features. As pretraining progresses, R2O follows a region-to-object curriculum which encourages learning region-level features early on and gradually progresses to train object-centric representations. Representations learned using R2O lead to state-of-the art performance in semantic segmentation for PASCAL VOC (+0.7 mIOU) and Cityscapes (+0.4 mIOU) and instance segmentation on MS COCO (+0.3 mask AP). Further, after pretraining on ImageNet, R2O pretrained models are able to surpass existing state-of-the-art in unsupervised object segmentation on the Caltech-UCSD Birds 200-2011 dataset (+2.9 mIoU) without any further training. We provide the code/models from this work at https://github.com/KKallidromitis/r2o.

Shufan Li, Konstantinos Kallidromitis, Akash Gokul et al.

World models have shown great utility in improving the task performance of embodied agents. While prior work largely focuses on pixel-space world models, these approaches face practical limitations in GUI settings, where predicting complex visual elements in future states is often difficult. In this work, we explore an alternative formulation of world modeling for GUI agents, where state transitions are described in natural language rather than predicting raw pixels. First, we introduce MobileWorldBench, a benchmark that evaluates the ability of vision-language models (VLMs) to function as world models for mobile GUI agents. Second, we release MobileWorld, a large-scale dataset consisting of 1.4M samples, that significantly improves the world modeling capabilities of VLMs. Finally, we propose a novel framework that integrates VLM world models into the planning framework of mobile agents, demonstrating that semantic world models can directly benefit mobile agents by improving task success rates. The code and dataset is available at https://github.com/jacklishufan/MobileWorld

Shungo Fujii, Yasunori Ishii, Kazuki Kozuka et al.

Data augmentation is an essential technique for improving recognition accuracy in object recognition using deep learning. Methods that generate mixed data from multiple data sets, such as mixup, can acquire new diversity that is not included in the training data, and thus contribute significantly to accuracy improvement. However, since the data selected for mixing are randomly sampled throughout the training process, there are cases where appropriate classes or data are not selected. In this study, we propose a data augmentation method that calculates the distance between classes based on class probabilities and can select data from suitable classes to be mixed in the training process. Mixture data is dynamically adjusted according to the training trend of each class to facilitate training. The proposed method is applied in combination with conventional methods for generating mixed data. Evaluation experiments show that the proposed method improves recognition performance on general and long-tailed image recognition datasets.

Hiroki Adachi, Tsubasa Hirakawa, Takayoshi Yamashita et al.

While convolutional neural networks (CNNs) have achieved excellent performances in various computer vision tasks, they often misclassify with malicious samples, a.k.a. adversarial examples. Adversarial training is a popular and straightforward technique to defend against the threat of adversarial examples. Unfortunately, CNNs must sacrifice the accuracy of standard samples to improve robustness against adversarial examples when adversarial training is used. In this work, we propose Masking and Mixing Adversarial Training (M2AT) to mitigate the trade-off between accuracy and robustness. We focus on creating diverse adversarial examples during training. Specifically, our approach consists of two processes: 1) masking a perturbation with a binary mask and 2) mixing two partially perturbed images. Experimental results on CIFAR-10 dataset demonstrate that our method achieves better robustness against several adversarial attacks than previous methods.

Risako Tanigawa, Yasunori Ishii, Kazuki Kozuka et al.

Color images are easy to understand visually and can acquire a great deal of information, such as color and texture. They are highly and widely used in tasks such as segmentation. On the other hand, in indoor person segmentation, it is necessary to collect person data considering privacy. We propose a new task for human segmentation from invisible information, especially airborne ultrasound. We first convert ultrasound waves to reflected ultrasound directional images (ultrasound images) to perform segmentation from invisible information. Although ultrasound images can roughly identify a person's location, the detailed shape is ambiguous. To address this problem, we propose a collaborative learning probabilistic U-Net that uses ultrasound and segmentation images simultaneously during training, closing the probabilistic distributions between ultrasound and segmentation images by comparing the parameters of the latent spaces. In inference, only ultrasound images can be used to obtain segmentation results. As a result of performance verification, the proposed method could estimate human segmentations more accurately than conventional probabilistic U-Net and other variational autoencoder models.

Risako Tanigawa, Yasunori Ishii, Kazuki Kozuka et al.

In action understanding in indoor, we have to recognize human pose and action considering privacy. Although camera images can be used for highly accurate human action recognition, camera images do not preserve privacy. Therefore, we propose a new task for human instance segmentation from invisible information, especially airborne ultrasound, for action recognition. To perform instance segmentation from invisible information, we first convert sound waves to reflected sound directional images (sound images). Although the sound images can roughly identify the location of a person, the detailed shape is ambiguous. To address this problem, we propose a collaborative learning variational autoencoder (CL-VAE) that simultaneously uses sound and RGB images during training. In inference, it is possible to obtain instance segmentation results only from sound images. As a result of performance verification, CL-VAE could estimate human instance segmentations more accurately than conventional variational autoencoder and some other models. Since this method can obtain human segmentations individually, it could be applied to human action recognition tasks with privacy protection.

Shufan Li, Konstantinos Kallidromitis, Akash Gokul et al.

We introduce OmniFlow, a novel generative model designed for any-to-any generation tasks such as text-to-image, text-to-audio, and audio-to-image synthesis. OmniFlow advances the rectified flow (RF) framework used in text-to-image models to handle the joint distribution of multiple modalities. It outperforms previous any-to-any models on a wide range of tasks, such as text-to-image and text-to-audio synthesis. Our work offers three key contributions: First, we extend RF to a multi-modal setting and introduce a novel guidance mechanism, enabling users to flexibly control the alignment between different modalities in the generated outputs. Second, we propose a novel architecture that extends the text-to-image MMDiT architecture of Stable Diffusion 3 and enables audio and text generation. The extended modules can be efficiently pretrained individually and merged with the vanilla text-to-image MMDiT for fine-tuning. Lastly, we conduct a comprehensive study on the design choices of rectified flow transformers for large-scale audio and text generation, providing valuable insights into optimizing performance across diverse modalities. The Code will be available at https://github.com/jacklishufan/OmniFlows.

Noriyuki Kugo, Xiang Li, Zixin Li et al.

Video Question Answering (VQA) inherently relies on multimodal reasoning, integrating visual, temporal, and linguistic cues to achieve a deeper understanding of video content. However, many existing methods rely on feeding frame-level captions into a single model, making it difficult to adequately capture temporal and interactive contexts. To address this limitation, we introduce VideoMultiAgents, a framework that integrates specialized agents for vision, scene graph analysis, and text processing. It enhances video understanding leveraging complementary multimodal reasoning from independently operating agents. Our approach is also supplemented with a question-guided caption generation, which produces captions that highlight objects, actions, and temporal transitions directly relevant to a given query, thus improving the answer accuracy. Experimental results demonstrate that our method achieves state-of-the-art performance on Intent-QA (79.0%, +6.2% over previous SOTA), EgoSchema subset (75.4%, +3.4%), and NExT-QA (79.6%, +0.4%). The source code is available at https://github.com/PanasonicConnect/VideoMultiAgents.

Zane Durante, Robathan Harries, Edward Vendrow et al.

Understanding Activities of Daily Living (ADLs) is a crucial step for different applications including assistive robots, smart homes, and healthcare. However, to date, few benchmarks and methods have focused on complex ADLs, especially those involving multi-person interactions in home environments. In this paper, we propose a new dataset and benchmark, InteractADL, for understanding complex ADLs that involve interaction between humans (and objects). Furthermore, complex ADLs occurring in home environments comprise a challenging long-tailed distribution due to the rarity of multi-person interactions, and pose fine-grained visual recognition tasks due to the presence of semantically and visually similar classes. To address these issues, we propose a novel method for fine-grained few-shot video classification called Name Tuning that enables greater semantic separability by learning optimal class name vectors. We show that Name Tuning can be combined with existing prompt tuning strategies to learn the entire input text (rather than only learning the prompt or class names) and demonstrate improved performance for few-shot classification on InteractADL and 4 other fine-grained visual classification benchmarks. For transparency and reproducibility, we release our code at https://github.com/zanedurante/vlm_benchmark.

Denis Gudovskiy, Shun Ishizaka, Kazuki Kozuka

Unsupervised anomaly detection with localization has many practical applications when labeling is infeasible and, moreover, when anomaly examples are completely missing in the train data. While recently proposed models for such data setup achieve high accuracy metrics, their complexity is a limiting factor for real-time processing. In this paper, we propose a real-time model and analytically derive its relationship to prior methods. Our CFLOW-AD model is based on a conditional normalizing flow framework adopted for anomaly detection with localization. In particular, CFLOW-AD consists of a discriminatively pretrained encoder followed by a multi-scale generative decoders where the latter explicitly estimate likelihood of the encoded features. Our approach results in a computationally and memory-efficient model: CFLOW-AD is faster and smaller by a factor of 10x than prior state-of-the-art with the same input setting. Our experiments on the MVTec dataset show that CFLOW-AD outperforms previous methods by 0.36% AUROC in detection task, by 1.12% AUROC and 2.5% AUPRO in localization task, respectively. We open-source our code with fully reproducible experiments.

Shufan Li, Konstantinos Kallidromitis, Akash Gokul et al.

We present Diffusion-KTO, a novel approach for aligning text-to-image diffusion models by formulating the alignment objective as the maximization of expected human utility. Since this objective applies to each generation independently, Diffusion-KTO does not require collecting costly pairwise preference data nor training a complex reward model. Instead, our objective requires simple per-image binary feedback signals, e.g. likes or dislikes, which are abundantly available. After fine-tuning using Diffusion-KTO, text-to-image diffusion models exhibit superior performance compared to existing techniques, including supervised fine-tuning and Diffusion-DPO, both in terms of human judgment and automatic evaluation metrics such as PickScore and ImageReward. Overall, Diffusion-KTO unlocks the potential of leveraging readily available per-image binary signals and broadens the applicability of aligning text-to-image diffusion models with human preferences.

Shufan Li, Konstantinos Kallidromitis, Hritik Bansal et al.

Modern Vision-Language Models (VLMs) can solve a wide range of tasks requiring visual reasoning. In real-world scenarios, desirable properties for VLMs include fast inference and controllable generation (e.g., constraining outputs to adhere to a desired format). However, existing autoregressive (AR) VLMs like LLaVA struggle in these aspects. Discrete diffusion models (DMs) offer a promising alternative, enabling parallel decoding for faster inference and bidirectional context for controllable generation through text-infilling. While effective in language-only settings, DMs' potential for multimodal tasks is underexplored. We introduce LaViDa, a family of VLMs built on DMs. We build LaViDa by equipping DMs with a vision encoder and jointly fine-tune the combined parts for multimodal instruction following. To address challenges encountered, LaViDa incorporates novel techniques such as complementary masking for effective training, prefix KV cache for efficient inference, and timestep shifting for high-quality sampling. Experiments show that LaViDa achieves competitive or superior performance to AR VLMs on multi-modal benchmarks such as MMMU, while offering unique advantages of DMs, including flexible speed-quality tradeoff, controllability, and bidirectional reasoning. On COCO captioning, LaViDa surpasses Open-LLaVa-Next-8B by +4.1 CIDEr with 1.92x speedup. On bidirectional tasks, it achieves +59% improvement on Constrained Poem Completion. These results demonstrate LaViDa as a strong alternative to AR VLMs. Code and models will be released in the camera-ready version.

Shufan Li, Konstantinos Kallidromitis, Akash Gokul et al.

The predominant approach to advancing text-to-image generation has been training-time scaling, where larger models are trained on more data using greater computational resources. While effective, this approach is computationally expensive, leading to growing interest in inference-time scaling to improve performance. Currently, inference-time scaling for text-to-image diffusion models is largely limited to best-of-N sampling, where multiple images are generated per prompt and a selection model chooses the best output. Inspired by the recent success of reasoning models like DeepSeek-R1 in the language domain, we introduce an alternative to naive best-of-N sampling by equipping text-to-image Diffusion Transformers with in-context reflection capabilities. We propose Reflect-DiT, a method that enables Diffusion Transformers to refine their generations using in-context examples of previously generated images alongside textual feedback describing necessary improvements. Instead of passively relying on random sampling and hoping for a better result in a future generation, Reflect-DiT explicitly tailors its generations to address specific aspects requiring enhancement. Experimental results demonstrate that Reflect-DiT improves performance on the GenEval benchmark (+0.19) using SANA-1.0-1.6B as a base model. Additionally, it achieves a new state-of-the-art score of 0.81 on GenEval while generating only 20 samples per prompt, surpassing the previous best score of 0.80, which was obtained using a significantly larger model (SANA-1.5-4.8B) with 2048 samples under the best-of-N approach.

XuDong Wang, Shaolun Zhang, Shufan Li et al.

We present SegLLM, a novel multi-round interactive reasoning segmentation model that enhances LLM-based segmentation by exploiting conversational memory of both visual and textual outputs. By leveraging a mask-aware multimodal LLM, SegLLM re-integrates previous segmentation results into its input stream, enabling it to reason about complex user intentions and segment objects in relation to previously identified entities, including positional, interactional, and hierarchical relationships, across multiple interactions. This capability allows SegLLM to respond to visual and text queries in a chat-like manner. Evaluated on the newly curated MRSeg benchmark, SegLLM outperforms existing methods in multi-round interactive reasoning segmentation by over 20%. Additionally, we observed that training on multi-round reasoning segmentation data enhances performance on standard single-round referring segmentation and localization tasks, resulting in a 5.5% increase in cIoU for referring expression segmentation and a 4.5% improvement in Acc@0.5 for referring expression localization.

Tiange Xiang, Adam Sun, Scott Delp et al.

Rendering the visual appearance of moving humans from occluded monocular videos is a challenging task. Most existing research renders 3D humans under ideal conditions, requiring a clear and unobstructed scene. Those methods cannot be used to render humans in real-world scenes where obstacles may block the camera's view and lead to partial occlusions. In this work, we present Wild2Avatar, a neural rendering approach catered for occluded in-the-wild monocular videos. We propose occlusion-aware scene parameterization for decoupling the scene into three parts - occlusion, human, and background. Additionally, extensive objective functions are designed to help enforce the decoupling of the human from both the occlusion and the background and to ensure the completeness of the human model. We verify the effectiveness of our approach with experiments on in-the-wild videos.

Chaitanya Patel, Hiroki Nakamura, Yuta Kyuragi et al. · salesforce, stanford

Egocentric human motion generation and forecasting with scene-context is crucial for enhancing AR/VR experiences, improving human-robot interaction, advancing assistive technologies, and enabling adaptive healthcare solutions by accurately predicting and simulating movement from a first-person perspective. However, existing methods primarily focus on third-person motion synthesis with structured 3D scene contexts, limiting their effectiveness in real-world egocentric settings where limited field of view, frequent occlusions, and dynamic cameras hinder scene perception. To bridge this gap, we introduce Egocentric Motion Generation and Egocentric Motion Forecasting, two novel tasks that utilize first-person images for scene-aware motion synthesis without relying on explicit 3D scene. We propose UniEgoMotion, a unified conditional motion diffusion model with a novel head-centric motion representation tailored for egocentric devices. UniEgoMotion's simple yet effective design supports egocentric motion reconstruction, forecasting, and generation from first-person visual inputs in a unified framework. Unlike previous works that overlook scene semantics, our model effectively extracts image-based scene context to infer plausible 3D motion. To facilitate training, we introduce EE4D-Motion, a large-scale dataset derived from EgoExo4D, augmented with pseudo-ground-truth 3D motion annotations. UniEgoMotion achieves state-of-the-art performance in egocentric motion reconstruction and is the first to generate motion from a single egocentric image. Extensive evaluations demonstrate the effectiveness of our unified framework, setting a new benchmark for egocentric motion modeling and unlocking new possibilities for egocentric applications.

Konstantinos Kallidromitis, Denis Gudovskiy, Kazuki Kozuka et al.

Recent contrastive methods show significant improvement in self-supervised learning in several domains. In particular, contrastive methods are most effective where data augmentation can be easily constructed e.g. in computer vision. However, they are less successful in domains without established data transformations such as time series data. In this paper, we propose a novel self-supervised learning framework that combines contrastive learning with neural processes. It relies on recent advances in neural processes to perform time series forecasting. This allows to generate augmented versions of data by employing a set of various sampling functions and, hence, avoid manually designed augmentations. We extend conventional neural processes and propose a new contrastive loss to learn times series representations in a self-supervised setup. Therefore, unlike previous self-supervised methods, our augmentation pipeline is task-agnostic, enabling our method to perform well across various applications. In particular, a ResNet with a linear classifier trained using our approach is able to outperform state-of-the-art techniques across industrial, medical and audio datasets improving accuracy over 10% in ECG periodic data. We further demonstrate that our self-supervised representations are more efficient in the latent space, improving multiple clustering indexes and that fine-tuning our method on 10% of labels achieves results competitive to fully-supervised learning.

Nishant Rai, Haofeng Chen, Jingwei Ji et al.

Existing research on action recognition treats activities as monolithic events occurring in videos. Recently, the benefits of formulating actions as a combination of atomic-actions have shown promise in improving action understanding with the emergence of datasets containing such annotations, allowing us to learn representations capturing this information. However, there remains a lack of studies that extend action composition and leverage multiple viewpoints and multiple modalities of data for representation learning. To promote research in this direction, we introduce Home Action Genome (HOMAGE): a multi-view action dataset with multiple modalities and view-points supplemented with hierarchical activity and atomic action labels together with dense scene composition labels. Leveraging rich multi-modal and multi-view settings, we propose Cooperative Compositional Action Understanding (CCAU), a cooperative learning framework for hierarchical action recognition that is aware of compositional action elements. CCAU shows consistent performance improvements across all modalities. Furthermore, we demonstrate the utility of co-learning compositions in few-shot action recognition by achieving 28.6% mAP with just a single sample.

Denis Gudovskiy, Luca Rigazio, Shun Ishizaka et al.

AutoAugment has sparked an interest in automated augmentation methods for deep learning models. These methods estimate image transformation policies for train data that improve generalization to test data. While recent papers evolved in the direction of decreasing policy search complexity, we show that those methods are not robust when applied to biased and noisy data. To overcome these limitations, we reformulate AutoAugment as a generalized automated dataset optimization (AutoDO) task that minimizes the distribution shift between test data and distorted train dataset. In our AutoDO model, we explicitly estimate a set of per-point hyperparameters to flexibly change distribution of train data. In particular, we include hyperparameters for augmentation, loss weights, and soft-labels that are jointly estimated using implicit differentiation. We develop a theoretical probabilistic interpretation of this framework using Fisher information and show that its complexity scales linearly with the dataset size. Our experiments on SVHN, CIFAR-10/100, and ImageNet classification show up to 9.3% improvement for biased datasets with label noise compared to prior methods and, importantly, up to 36.6% gain for underrepresented SVHN classes.