Aniket Rege, Arka Sadhu, Yuliang Li, Kejie Li, Ramya Korlakai Vinayak, Yuning Chai, Yong Jae Lee, Hyo Jin Kim. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Active Speaker Detection (ASD) aims to identify who is speaking in complex visual scenes. While humans naturally rely on lip-audio synchronization, existing ASD models often misclassify non-speaking instances when lip movements and audio are unsynchronized. To address this, we propose Lip landmark Assisted Speaker dEtection for Robustness (LASER), which explicitly incorporates lip landmarks during training to guide the model’s attention to speech-relevant regions. Given a face track, LASER extracts visual features and encodes 2D lip landmarks into dense maps. To handle failure cases such as low resolution or occlusion, we introduce an auxiliary consistency loss that aligns lip-aware and face-only predictions, removing the need for landmark detectors at test time. LASER outperforms state-of-the-art models across both in-domain and out-of-domain benchmarks. To further evaluate robustness in realistic conditions, we introduce LASER-bench, a curated dataset of modern video clips with varying levels of background noise. On the high-noise subset, LASER improves mAP by 3.3 and 4.3 points over LoCoNet and TalkNet, respectively, demonstrating strong resilience to real-world acoustic challenges.

Long-form video understanding with Large Vision Language Models is challenged by the need to analyze temporally dispersed yet spatially concentrated key moments within limited context windows. In this work, we introduce VideoMindPalace, a new framework inspired by the "Mind Palace", which organizes critical video moments into a topologically structured semantic graph. VideoMindPalace organizes key information through (i) hand-object tracking and interaction, (ii) clustered activity zones representing specific areas of recurring activities, and (iii) environment layout mapping, allowing natural language parsing by LLMs to provide grounded insights on spatio-temporal and 3D context. In addition, we propose the Video Mind-Palace Benchmark (VMB), to assess human-like reasoning, including spatial localization, temporal reasoning, and layout-aware sequential understanding. Evaluated on VMB and established video QA datasets, including EgoSchema, NExT-QA, IntentQA, and the Active Memories Benchmark, VideoMindPalace demonstrates notable gains in spatiotemporal coherence and human-aligned reasoning, advancing long-form video analysis capabilities in VLMs. Project website: https://oodbag.github.io/VMP/.

Large Multimodal Models (e.g., GPT-4, Gemini, Chameleon) have evolved into powerful tools with millions of users. However, they remain generic models and lack personalized knowledge of specific user concepts. Previous work has explored personalization for text generation, yet it remains unclear how these methods can be adapted to new modalities, such as image generation. In this paper, we introduce Yo’Chameleon, the first attempt to study personalization for large multimodal models. Given 3-5 images of a particular concept, Yo’Chameleon leverages soft-prompt tuning to embed subject-specific information to (i) answer questions about the subject and (ii) recreate pixel-level details to produce images of the subject in new contexts. Yo’Chameleon is trained with (i) a self-prompting optimization mechanism to balance performance across multiple modalities, and (ii) a "soft-positive" image generation approach to enhance image quality in a few-shot setting. Our qualitative and quantitative analyses reveal that Yo’Chameleon can learn concepts more efficiently using fewer tokens and effectively encode visual attributes, outperforming prompting baselines.

We propose X-Fusion, a framework that extends pretrained Large Language Models (LLMs) for multimodal tasks while preserving their language capabilities. X-Fusion employs a dual-tower design with modality-specific weights, keeping the LLM's parameters frozen while integrating vision-specific information for both understanding and generation. Our experiments demonstrate that X-Fusion consistently outperforms alternative architectures on both image-to-text and text-to-image tasks. We find that incorporating understanding-focused data improves generation quality, reducing image data noise enhances overall performance, and feature alignment accelerates convergence for smaller models but has minimal impact on larger ones. Our findings provide valuable insights into building efficient unified multimodal models.

Popular text-to-image (T2I) systems are trained on web-scraped data, which is heavily Amero and Euro-centric, underrepresenting the cultures of the Global South. To analyze these biases, we introduce CuRe, a novel and scalable benchmarking and scoring suite for cultural representativeness that leverages the marginal utility of attribute specification to T2I systems as a proxy for human judgments. Our CuRe benchmark dataset has a novel categorical hierarchy built from the crowdsourced Wikimedia knowledge graph, with 300 cultural artifacts across 32 cultural subcategories grouped into six broad cultural axes (food, art, fashion, architecture, celebrations, and people). Our dataset's categorical hierarchy enables CuRe scorers to evaluate T2I systems by analyzing their response to increasing the informativeness of text conditioning, enabling fine-grained cultural comparisons. We empirically observe much stronger correlations of our class of scorers to human judgments of perceptual similarity, image-text alignment, and cultural diversity across image encoders (SigLIP 2, AIMV2 and DINOv2), vision-language models (OpenCLIP, SigLIP 2, Gemini 2.0 Flash) and state-of-the-art text-to-image systems, including three variants of Stable Diffusion (1.5, XL, 3.5 Large), FLUX.1 [dev], Ideogram 2.0, and DALL-E 3. The code and dataset is open-sourced and available at https://aniketrege.github.io/cure/.

Large Multimodal Models (LMMs) have shown significant visual reasoning capabilities by connecting a visual encoder and a large language model. LMMs typically take in a fixed and large amount of visual tokens, such as the penultimate layer features in the CLIP visual encoder, as the prefix content. Recent LMMs incorporate more complex visual inputs, such as high-resolution images and videos, which further increases the number of visual tokens significantly. However, due to the inherent design of the Transformer architecture, the computational costs of these models tend to increase quadratically with the number of input tokens. To tackle this problem, we explore a token reduction mechanism that identifies significant spatial redundancy among visual tokens. In response, we propose PruMerge, a novel adaptive visual token reduction strategy that significantly reduces the number of visual tokens without compromising the performance of LMMs. Specifically, to metric the importance of each token, we exploit the sparsity observed in the visual encoder, characterized by the sparse distribution of attention scores between the class token and visual tokens. This sparsity enables us to dynamically select the most crucial visual tokens to retain. Subsequently, we cluster the selected (unpruned) tokens based on their key similarity and merge them with the unpruned tokens, effectively supplementing and enhancing their informational content. Empirically, when applied to LLaVA-1.5, our approach can compress the visual tokens by 14 times on average, and achieve comparable performance across diverse visual question-answering and reasoning tasks. Code and checkpoints are at https://llava-prumerge.github.io/.

Vision Language Models (VLMs) have recently been leveraged to generate robotic actions, forming Vision-Language-Action (VLA) models. However, directly adapting a pretrained VLM for robotic control remains challenging, particularly when constrained by a limited number of robot demonstrations. In this work, we introduce LLaRA: Large Language and Robotics Assistant, a framework that formulates robot action policy as visuo-textual conversations and enables an efficient transfer of a pretrained VLM into a powerful VLA, motivated by the success of visual instruction tuning in Computer Vision. First, we present an automated pipeline to generate conversation-style instruction tuning data for robots from existing behavior cloning datasets, aligning robotic actions with image pixel coordinates. Further, we enhance this dataset in a self-supervised manner by defining six auxiliary tasks, without requiring any additional action annotations. We show that a VLM finetuned with a limited amount of such datasets can produce meaningful action decisions for robotic control. Through experiments across multiple simulated and real-world tasks, we demonstrate that LLaRA achieves state-of-the-art performance while preserving the generalization capabilities of large language models. The code, datasets, and pretrained models are available at https://github.com/LostXine/LLaRA.

Large Multimodal Models (LMMs) such as LLaVA have shown strong performance in visual-linguistic reasoning. These models first embed images into a fixed large number of visual tokens and then feed them into a Large Language Model (LLM). However, this design causes an excessive number of tokens for dense visual scenarios such as high-resolution images and videos, leading to great inefficiency. While token pruning/merging methods do exist, they produce a single length output for each image and do not afford flexibility in trading off information density v.s. efficiency. Inspired by the concept of Matryoshka Dolls, we propose : Matryoshka Multimodal Models, which learns to represent visual content as nested sets of visual tokens that capture information across multiple coarse-to-fine granularities. Our approach offers several unique benefits for LMMs: (1) One can explicitly control the visual granularity per test instance during inference, e.g. , adjusting the number of tokens used to represent an image based on the anticipated complexity or simplicity of the content; (2) provides a framework for analyzing the granularity needed for existing datasets, where we find that COCO-style benchmarks only need around 9 visual tokens to obtain accuracy similar to that of using all 576 tokens; (3) Our approach provides a foundation to explore the best trade-off between performance and visual token length at sample level, where our investigation reveals that a large gap exists between the oracle upper bound and current fixed-scale representations.

As latent diffusion models (LDMs) democratize image generation capabilities, there is a growing need to detect fake images. A good detector should focus on the generative model’s fingerprints while ignoring image properties such as semantic content, resolution, file format, etc. Fake image detectors are usually built in a data-driven way, where a model is trained to separate real from fake images. Existing works primarily investigate network architecture choices and training recipes. In this work, we argue that in addition to these algorithmic choices, we also require a well-aligned dataset of real/fake images to train a robust detector. For the family of LDMs, we propose a very simple way to achieve this: we reconstruct all the real images using the LDM's autoencoder, without any denoising operation. We then train a model to separate these real images from their reconstructions. The fakes created this way are extremely similar to the real ones in almost every aspect (e.g., size, aspect ratio, semantic content), which forces the model to look for the LDM decoder's artifacts. We empirically show that this way of creating aligned real/fake datasets, which also sidesteps the computationally expensive denoising process, helps in building a detector that focuses less on spurious correlations, something that a very popular existing method is susceptible to. Finally, to demonstrate the effectivenss of dataset alignment, we build a detector using images that are not natural objects, and present promising results. Overall, our work identifies the subtle but significant issues that arise when training a fake image detector and proposes a simple and inexpensive solution to address these problems.

Detecting AI-generated images is a challenging yet essential task. A primary difficulty arises from the detector’s tendency to rely on spurious patterns, such as compression artifacts, which can influence its decisions. These issues often stem from specific patterns that the detector associates with the real data distribution, making it difficult to isolate the actual generative traces. We argue that an image should be classified as fake if and only if it contains artifacts introduced by the generative model. Based on this premise, we propose Stay-Positive, an algorithm designed to constrain the detector’s focus to generative artifacts while disregarding those associated with real data. Experimental results demonstrate that detectors trained with Stay-Positive exhibit reduced susceptibility to spurious correlations, leading to improved generalization and robustness to post-processing. Additionally, unlike detectors that associate artifacts with real images, those that focus purely on fake artifacts are better at detecting inpainted real images.

Multi-frame temporal inputs are important for vision-based autonomous driving. Observations from different angles enable the recovery of 3 D object states from 2 D images as long as we can identify the same instance from different input frames. However, the dynamic nature of driving scenes leads to significant variance in the instance appearance and shape captured by the cameras at different time steps. To this end, we propose a novel contrastive learning algorithm, Cohere3D, to learn coherent instance representations robust to the changes of distance and perspective in a long-term temporal sequence without any human annotations. In the pretraining stage, raw point clouds from LiDAR sensors are utilized to construct the instance-wise long-term temporal correspondence, which serves as guidance for the extraction of instance-level representation from the vision-based bird's-eye-view (BEV) feature map. Cohere3D encourages consistent representation for the same instance at different frames but distinguishes between different instances. We validate the effectiveness and generalizability of our algorithm by finetuning the pretrained model across key downstream autonomous driving tasks: perception, mapping, prediction, and planning. Results show a notable improvement in both data efficiency and final performance in all these tasks.

Large language models (LLMs) have made significant advancements in natural language understanding. However, through that enormous semantic representation that the LLM has learnt, is it somehow possible for it to understand images as well? This work investigates this question. To enable the LLM to process images, we convert them into a representation given by Scalable Vector Graphics (SVG). To study what the LLM can do with this XML-based textual description of images, we test the LLM on three broad computer vision tasks: (i) visual reasoning and question answering, (ii) image classification under distribution shift, few-shot learning, and (iii) generating new images using visual prompting. Even though we do not naturally associate LLMs with any visual understanding capabilities, our results indicate that the LLM can often do a decent job in many of these tasks, potentially opening new avenues for research into LLMs' ability to understand image data. Our code, data, and models can be found here https://github.com/mu-cai/svg-11m.

Advancements in large pre-trained generative models have expanded their potential as effective data generators in visual recognition. This work delves into the impact of generative images, primarily comparing paradigms that harness external data (i.e. generative vs. retrieval vs. original). Our key contributions are: 1) GenBench Construction: We devise GenBench, a broad benchmark comprising 22 datasets with 2548 categories, to appraise generative data across various visual recognition tasks. 2) CLER Score: To address the insufficient correlation of existing metrics (e.g., FID, CLIP score) with downstream recognition performance, we propose CLER, a training-free metric indicating generative data's efficiency for recognition tasks prior to training. 3) New Baselines: Comparisons of generative data with retrieved data from the same external pool help to elucidate the unique traits of generative data. 4) External Knowledge Injection: By fine-tuning special token embeddings for each category via Textual Inversion, performance improves across 17 datasets, except when dealing with low-resolution reference images. Our exhaustive benchmark and analysis spotlight generative data's promise in visual recognition, while identifying key challenges for future investigation.

While existing large vision-language multimodal models focus on whole image understanding, there is a prominent gap in achieving region-specific comprehension. Current approaches that use textual coordinates or spatial encodings often fail to provide a user-friendly interface for visual prompting. To address this challenge, we introduce a novel multimodal model capable of decoding arbitrary (free-form) visual prompts. This allows users to intuitively mark images and interact with the model using natural cues like a “red bounding box” or “pointed arrow”. Our simple design directly overlays visual markers onto the RGB image, eliminating the need for complex region encodings, yet achieves state-of-the-art performance on region-understanding tasks like Visual7W, PointQA, and Visual Commonsense Reasoning benchmark. Furthermore, we present ViP-Bench, a comprehensive benchmark to assess the capability of models in understanding visual prompts across multiple dimensions, enabling future research in this domain. Code, data, and model are publicly available.

Large multimodal models (LMM) have recently shown encouraging progress with visual instruction tuning. In this paper, we present the first systematic study to investigate the design choices of LMMs in a controlled setting under the LLaVA framework. We show that the fully-connected vision-language connector in LLaVA is surprisingly power-ful and data-efficient. With simple modifications to LLa VA, namely, using CLIP- ViT-L-336px with an MLP projection and adding academic-task-oriented VQA data with response formatting prompts, we establish stronger baselines that achieve state-of-the-art across 11 benchmarks. Our final 13B checkpoint uses merely 1.2M publicly available data, and finishes full training in ~ 1 day on a single 8-AI00 node. Furthermore, we present some early exploration of open problems in LMMs, including scaling to higher resolution inputs, compositional capabilities, and model hallucination, etc. We hope this makes state-of-the-art LMM research more accessible. Code and model will be publicly available.

In recent years, image editing has advanced remarkably. With increased human control, it is now possible to edit an image in a plethora of ways; from specifying in text what we want to change, to straight up dragging the contents of the image in an interactive point-based manner. However, most of the focus has remained on editing single images at a time. Whether and how we can simultaneously edit large batches of images has remained understudied. With the goal ofmin-imizing human supervision in the editing process, this paper presents a novel method for interactive batch image editing using StyleGAN as the medium. Given an edit specified by users in an example image (e.g., make the face frontal), our method can automatically transfer that edit to other test images, so that regardless of their initial state (pose), they all arrive at the same final state (e.g., all facing front). Ex-tensive experiments demonstrate that edits performed using our method have similar visual quality to existing single-image-editing methods, while having more visual consis-tency and saving significant time and human effort.

In the realm of vision models, the primary mode of representation is using pixels to rasterize the visual world.Yet this is not always the best or unique way to represent visual content, especially for designers and artists who depict the world using geometry primitives such as polygons.Vector graphics (VG), on the other hand, offer a textual representation of visual content, which can be more concise and powerful for content like cartoons, sketches and scientific figures.Recent studies have shown promising results on processing vector graphics with capable Large Language Models (LLMs).However, such works focus solely on qualitative results, understanding, or a specific type of vector graphics.We propose VGBench, a comprehensive benchmark for LLMs on handling vector graphics through diverse aspects, including (a) both visual understanding and generation, (b) evaluation of various vector graphics formats, (c) diverse question types, (d) wide range of prompting techniques, (e) under multiple LLMs and (f) comparison with VLMs on rasterized representations.Evaluating on our collected 4279 understanding and 5845 generation samples, we find that LLMs show strong capability on both aspects while exhibiting less desirable performance on low-level formats (SVG).Both data and evaluation pipeline will be open-sourced at https://vgbench.github.io.

3D perception in LiDAR point clouds is crucial for a self-driving vehicle to properly act in 3D environment. However, manually labeling point clouds is hard and costly. There has been a growing interest in self-supervised pre-training of 3D perception models. Following the success of contrastive learning in images, current methods mostly conduct contrastive pre-training on point clouds only. Yet an autonomous driving vehicle is typically supplied with multiple sensors including cameras and LiDAR. In this context, we systematically study single modality, cross-modality, and multi-modality for contrastive learning of point clouds, and show that cross-modality wins over other alternatives. In addition, considering the huge difference between the training sources in 2D images and 3D point clouds, it remains unclear how to design more effective contrastive units for LiDAR. We therefore propose the instance-aware and similarity-balanced contrastive units that are tailored for self-driving point clouds. Extensive experiments reveal that our approach achieves remarkable performance gains over various point cloud models across the downstream perception tasks of LiDAR based 3D object detection and 3D semantic segmentation on the four popular benchmarks including Waymo Open Dataset, nuScenes, SemanticKITTI and ONCE.

Large Multimodal Models (LMMs) have shown remarkable capabilities across a variety of tasks (e.g., image captioning, visual question answering). While broad, their knowledge remains generic (e.g., recognizing a dog), and they are unable to handle personalized subjects (e.g., recognizing a user's pet dog). Human reasoning, in contrast, typically operates within the context of specific subjects in our surroundings. For example, one might ask, "What should I buy for *my dog*'s birthday?"; as opposed to a generic inquiry about "What should I buy for *a dog*'s birthday?". Similarly, when looking at a friend's image, the interest lies in seeing their activities (e.g., "*my friend* is holding a cat"), rather than merely observing generic human actions (e.g., "*a man* is holding a cat"). In this paper, we introduce the novel task of personalizing LMMs, so that they can have conversations about a specific subject. We propose Yo'LLaVA, which learns to embed a personalized subject into a set of latent tokens given a handful of example images of the subject. Our qualitative and quantitative analyses reveal that Yo'LLaVA can learn the concept more efficiently using fewer tokens and more effectively encode the visual attributes compared to strong prompting baselines (e.g., LLaVA).

Foundation models possess strong capabilities in reasoning and memorizing across modalities. To further unleash the power of foundation models, we present FIND, a generalized interface for aligning foundation models' embeddings with unified image and dataset-level understanding spanning modality and granularity. As shown in Fig.1, a lightweight transformer interface without tuning any foundation model weights is enough for segmentation, grounding, and retrieval in an interleaved manner. The proposed interface has the following favorable attributes: (1) Generalizable. It applies to various tasks spanning retrieval, segmentation, etc., under the same architecture and weights. (2) Interleavable. With the benefit of multi-task multi-modal training, the proposed interface creates an interleaved shared embedding space. (3) Extendable. The proposed interface is adaptive to new tasks, and new models. In light of the interleaved embedding space, we introduce FIND-Bench, which introduces new training and evaluation annotations to the COCO dataset for interleaved segmentation and retrieval. We are the first work aligning foundations models' embeddings for interleave understanding. Meanwhile, our approach achieves state-of-the-art performance on FIND-Bench and competitive performance on standard retrieval and segmentation settings.

In precision livestock farming, the individual identification of cattle is crucial to inform the decisions made to enhance animal welfare, health, and productivity. In literature, models exist that can read ear tags; however, they are not easily portable to real-world cattle production environments and make predictions mainly on still images. We propose a video-based cattle ear tag reading system, called ReadMyCow, which takes advantage of the temporal characteristics in videos to accurately detect, track, and read cattle ear tags at 25 FPS on edge devices. For each frame in a video, ReadMyCow functions in two steps. 1) Tag detection: a YOLOv5s Object Detection model and NVIDIA Deepstream Tracking Layer detect and track the tags present. 2) Tag reading: the novel WhenToRead module decides whether to read each tag, using a TRBA Scene Text Recognition model, or to use the reading from a previous frame. The system is implemented on an edge device, namely the NVIDIA Jetson AGX Orin or Xavier, making it portable to cattle production environments without external computational resources. To attain real-time speeds, ReadMyCow only reads the detected tag in the current frame if it thinks it will get a better reading when a decision metric is significantly improved in the current frame. Ideally, this means the best reading of a tag is found and stored throughout a tag’s presence in the video, even when the tag becomes occluded or blurry. While testing the system at a real Midwestern dairy farm housing 9,000 cows, 96.1% of printed ear tags were accurately read by the ReadMyCow system, demonstrating its real-world commercial potential. ReadMyCow opens opportunities for informed data-driven decision-making processes on commercial cattle farms.

Large-scale text-to-image diffusion models have made amazing advances. However, the status quo is to use text input alone, which can impede controllability. In this work, we propose GLIGEN, Grounded-Language-to-Image Generation, a novel approach that builds upon and extends the functionality of existing pre-trained text-to-image diffusion models by enabling them to also be conditioned on grounding inputs. To preserve the vast concept knowledge of the pre-trained model, we freeze all of its weights and inject the grounding information into new trainable layers via a gated mechanism. Our model achieves open-world grounded text2img generation with caption and bounding box condition inputs, and the grounding ability generalizes well to novel spatial configurations and concepts. GLIGEN's zero-shot performance on COCO and LVIS outperforms existing supervised layout-to-image baselines by a large margin.

Image-text contrastive learning models such as CLIP have demonstrated strong task transfer ability. The high generality and usability of these visual models is achieved via a web-scale data collection process to ensure broad concept coverage, followed by expensive pre-training to feed all the knowledge into model weights. Alternatively, we propose React,REtrieval-Augmented CusTomization, a framework to acquire the relevant web knowledge to build customized visual models for target domains. We retrieve the most relevant image-text pairs$(\thicksim3\%$of CLIP pre-training data) from the web-scale database as external knowledge and propose to customize the model by only training new modularized blocks while freezing all the original weights. The effectiveness of Reactis demonstrated via extensive experiments on classification, retrieval, detection and segmentation tasks, including zero, few, and full-shot settings. Particularly, on the zero-shot classification task, compared with CLIP, it achieves up to 5.4% improvement on ImageNet and 3.7% on the Elevaterbenchmark (20 datasets).

With generative models proliferating at a rapid rate, there is a growing need for general purpose fake image detectors. In this work, we first show that the existing paradigm, which consists of training a deep network for real-vs-fake classification, fails to detect fake images from newer breeds of generative models when trained to detect GAN fake images. Upon analysis, we find that the resulting classifier is asymmetrically tuned to detect patterns that make an image fake. The real class becomes a ‘sink’ class holding anything that is not fake, including generated images from models not accessible during training. Building upon this discovery, we propose to perform real-vs-fake classification without learning; i.e., using a feature space not explicitly trained to distinguish real from fake images. We use nearest neighbor and linear probing as instantiations of this idea. When given access to the feature space of a large pretrained vision-language model, the very simple baseline of nearest neighbor classification has surprisingly good generalization ability in detecting fake images from a wide variety of generative models; e.g., it improves upon the SoTA [50] by +15.07 mAP and +25.90% acc when tested on unseen diffusion and autoregressive models. Our code, models, and data can be found at https://github.com/Yuheng-Li/UniversalFakeDetect

We present X-Decoder, a generalized decoding model that can predict pixel-level segmentation and language tokens seamlessly. X-Decoder takes as input two types of queries: (i) generic non-semantic queries and (ii) semantic queries induced from text inputs, to decode different pixel-level and token-level outputs in the same semantic space. With such a novel design, X-Decoder is the first work that provides a unified way to support all types of image segmentation and a variety of vision-language (VL) tasks. Without any pseudo-labeling, our design enables seamless interactions across tasks at different granularities and brings mutual benefits by learning a common and rich pixel-level understanding. After pretraining on a mixed set of a limited amount of segmentation data and millions of image-text pairs, X-Decoder exhibits strong transferability to a wide range of downstream tasks in both zero-shot and finetuning settings. Notably, it achieves (1) state-of-the-art results on open-vocabulary segmentation and referring segmentation on seven datasets; (2) better or competitive finetuned performance to other generalist and specialist models on segmentation and VL tasks; and (3) flexibility for efficient fine-tuning and novel task composition (e.g., referring captioning and image editing shown in Fig. 1). Code, demo, video and visualization are available at: https://x-decoder-vl.github.io.

Domain generalization studies the problem of training a model with samples from several domains (or distributions) and then testing the model with samples from a new, unseen domain. In this paper, we propose a novel approach for domain generalization that leverages recent advances in large vision-language models, specifically a CLIP teacher model, to train a smaller model that generalizes to unseen domains. The key technical contribution is a new type of regularization that requires the student’s learned image representations to be close to the teacher’s learned text representations obtained from encoding the corresponding text descriptions of images. We introduce two designs of the loss function, absolute and relative distance, which provide specific guidance on how the training process of the student model should be regularized. We evaluate our proposed method, dubbed RISE (Regularized Invariance with Semantic Embeddings), on various benchmark datasets, and show that it outperforms several state-of-the-art domain generalization methods. To our knowledge, our work is the first to leverage knowledge distillation using a large vision-language model for domain generalization. By incorporating text-based information, RISE improves the generalization capability of machine learning models.

Instruction tuning large language models (LLMs) using machine-generated instruction-following data has been shown to improve zero-shot capabilities on new tasks, but the idea is less explored in the multimodal field. We present the first attempt to use language-only GPT-4 to generate multimodal language-image instruction-following data. By instruction tuning on such generated data, we introduce LLaVA: Large Language and Vision Assistant, an end-to-end trained large multimodal model that connects a vision encoder and an LLM for general-purpose visual and language understanding. To facilitate future research on visual instruction following, we construct two evaluation benchmarks with diverse and challenging application-oriented tasks. Our experiments show that LLaVA demonstrates impressive multimodal chat abilities, sometimes exhibiting the behaviors of multimodal GPT-4 on unseen images/instructions, and yields a 85.1% relative score compared with GPT-4 on a synthetic multimodal instruction-following dataset. When fine-tuned on Science QA, the synergy of LLaVA and GPT-4 achieves a new state-of-the-art accuracy of 92.53%. We make GPT-4 generated visual instruction tuning data, our model, and code publicly available.

Text-conditioned image editing has emerged as a powerful tool for editing images. However, in many situations, language can be ambiguous and ineffective in describing specific image edits. When faced with such challenges, visual prompts can be a more informative and intuitive way to convey ideas. We present a method for image editing via visual prompting. Given pairs of example that represent the "before" and "after" images of an edit, our goal is to learn a text-based editing direction that can be used to perform the same edit on new images. We leverage the rich, pretrained editing capabilities of text-to-image diffusion models by inverting visual prompts into editing instructions. Our results show that with just one example pair, we can achieve competitive results compared to state-of-the-art text-conditioned image editing frameworks.

Knowledge distillation aims to transfer useful information from a teacher network to a student network, with the primary goal of improving the student's performance for the task at hand. Over the years, there has a been a deluge of novel techniques and use cases of knowledge distillation. Yet, despite the various improvements, there seems to be a glaring gap in the community's fundamental understanding of the process. Specifically, what is the knowledge that gets distilled in knowledge distillation? In other words, in what ways does the student become similar to the teacher? Does it start to localize objects in the same way? Does it get fooled by the same adversarial samples? Does its data invariance properties become similar? Our work presents a comprehensive study to try to answer these questions. We show that existing methods can indeed indirectly distill these properties beyond improving task performance. We further study why knowledge distillation might work this way, and show that our findings have practical implications as well.

In this work, we present SEEM, a promotable and interactive model for segmenting everything everywhere all at once in an image. In SEEM, we propose a novel and versatile decoding mechanism that enables diverse prompting for all types of segmentation tasks, aiming at a universal interface that behaves like large language models (LLMs). More specifically, SEEM is designed with four desiderata: i) Versatility. We introduce a new visual prompt to unify different spatial queries including points, boxes, scribbles, and masks, which can further generalize to a different referring image; ii) Compositionality. We learn a joint visual-semantic space between text and visual prompts, which facilitates the dynamic composition of two prompt types required for various segmentation tasks, as shown in Fig. 1; iii) Interactivity. We further incorporate learnable memory prompts into the decoder to retain segmentation history through mask-guided cross-attention from the decoder to image features; iv) Semantic awareness. We use a text encoder to encode text queries and mask labels into the same semantic space for open-vocabulary segmentation. We conduct a comprehensive empirical study to validate the effectiveness of SEEM across diverse segmentation tasks. The results demonstrate that SEEM exhibits robust generalizing to unseen user intents as it learns to compose prompts of different types in a unified representation space. Our approach achieves competitive performance on interactive segmentation, generic segmentation, referring segmentation, and video object segmentation on 9 datasets with minimum 1/100 supervision in a single set of weights.

Training with an emphasis on “hard-to-learn” components of the data has been proven as an effective method to improve the generalization of machine learning models, especially in the settings where robustness (e.g., generalization across distributions) is valued. Existing literature discussing this “hard-to-learn” concept are mainly expanded either along the dimension of the samples or the dimension of the features. In this paper, we aim to introduce a simple view merging these two dimensions, leading to a new, simple yet effective, heuristic to train machine learning models by emphasizing the worst-cases on both the sample and the feature dimensions. We name our method W2D following the concept of “Worst-case along Two Dimensions”. We validate the idea and demonstrate its empirical strength over standard benchmarks.

3D-aware generative models have shown that the introduction of 3D information can lead to more controllable image generation. In particular, the current state-of-the-art model GIRAFFE [38] can control each object's rotation, translation, scale, and scene camera pose without corresponding supervision. However, GIRAFFE only operates well when the image resolution is low. We propose GIRAFFE HD, a high-resolution 3D-aware generative model that inherits all of GIRAFFE's controllable features while generating high-quality, high-resolution images (5122resolution and above). The key idea is to leverage a style- based neural renderer, and to independently generate the foreground and background to force their disentanglement while imposing consistency constraints to stitch them together to composite a coherent final image. We demonstrate state-of-the-art 3D controllable high-resolution image generation on multiple natural image datasets.

Learning visual representations from natural language supervision has recently shown great promise in a number of pioneering works. In general, these language-augmented visual models demonstrate strong transferability to a variety of datasets/tasks. However, it remains challenging to evaluate the transferablity of these foundation models due to the lack of easy-to-use toolkits for fair benchmarking. To tackle this, we build ELEVATER (Evaluation of Language-augmented Visual Task-level Transfer), the first benchmark to compare and evaluate pre-trained language-augmented visual models. Several highlights include: (i) Datasets. As downstream evaluation suites, it consists of 20 image classification datasets and 35 object detection datasets, each of which is augmented with external knowledge. (ii) Toolkit. An automatic hyper-parameter tuning toolkit is developed to ensure the fairness in model adaption. To leverage the full power of language-augmented visual models, novel language-aware initialization methods are proposed to significantly improve the adaption performance. (iii) Metrics. A variety of evaluation metrics are used, including sample-efficiency (zero-shot and few-shot) and parameter-efficiency (linear probing and full model fine-tuning). We will publicly release ELEVATER.

Machine learning has demonstrated remarkable prediction accuracy over i.i.d data, but the accuracy often drops when tested with data from another distribution. In this paper, we aim to offer another view of this problem in a perspective assuming the reason behind this accuracy drop is the reliance of models on the features that are not aligned well with how a data annotator considers similar across these two datasets. We refer to these features as misaligned features. We extend the conventional generalization error bound to a new one for this setup with the knowledge of how the misaligned features are associated with the label. Our analysis offers a set of techniques for this problem, and these techniques are naturally linked to many previous methods in robust machine learning literature. We also compared the empirical strength of these methods demonstrated the performance when these previous techniques are combined, with implementation available.

Timely detection of horse pain is important for equine welfare. Horses express pain through their facial and body behavior, but may hide signs of pain from unfamiliar human observers. In addition, collecting visual data with detailed annotation of horse behavior and pain state is both cumbersome and not scalable. Consequently, a pragmatic equine pain classification system would use video of the unobserved horse and weak labels. This paper proposes such a method for equine pain classification by using multi-view surveillance video footage of unobserved horses with induced orthopaedic pain, with temporally sparse video level pain labels. To ensure that pain is learned from horse body language alone, we first train a self-supervised generative model to disentangle horse pose from its appearance and background before using the disentangled horse pose latent representation for pain classification. To make best use of the pain labels, we develop a novel loss that formulates pain classification as a multi-instance learning problem. Our method achieves pain classification accuracy better than human expert performance with 60% accuracy. The learned latent horse pose representation is shown to be viewpoint covariant, and disentangled from horse appearance. Qualitative analysis of pain classified segments shows correspondence between the pain symptoms identified by our model, and equine pain scales used in veterinary practice.

We present a simple, fully-convolutional model for real-time ( fps) instance segmentation that achieves competitive results on MS COCO evaluated on a single Titan Xp, which is significantly faster than any previous state-of-the-art approach. Moreover, we obtain this result after training on only one GPU. We accomplish this by breaking instance segmentation into two parallel subtasks: (1) generating a set of prototype masks and (2) predicting per-instance mask coefficients. Then we produce instance masks by linearly combining the prototypes with the mask coefficients. We find that because this process doesn't depend on repooling, this approach produces very high-quality masks and exhibits temporal stability for free. Furthermore, we analyze the emergent behavior of our prototypes and show they learn to localize instances on their own in a translation variant manner, despite being fully-convolutional. We also propose Fast NMS, a drop-in 12 ms faster replacement for standard NMS that only has a marginal performance penalty. Finally, by incorporating deformable convolutions into the backbone network, optimizing the prediction head with better anchor scales and aspect ratios, and adding a novel fast mask re-scoring branch, our YOLACT++ model can achieve 34.1 mAP on MS COCO at 33.5 fps, which is fairly close to the state-of-the-art approaches while still running at real-time.

Training generative models, such as GANs, on a target domain containing limited examples (e.g., 10) can easily result in overfitting. In this work, we seek to utilize a large source domain for pretraining and transfer the diversity information from source to target. We propose to preserve the relative similarities and differences between instances in the source via a novel cross-domain distance consistency loss. To further reduce overfitting, we present an anchor-based strategy to encourage different levels of realism over different regions in the latent space. With extensive results in both photorealistic and non-photorealistic domains, we demonstrate qualitatively and quantitatively that our few-shot model automatically discovers correspondences between source and target domains and generates more diverse and realistic images than previous methods.

Video inpainting aims to fill spatiotemporal "corrupted" regions with plausible content. To achieve this goal, it is necessary to find correspondences from neighbouring frames to faithfully hallucinate the unknown con-tent. Current methods achieve this goal through attention, flow-based warping, or 3D temporal convolution. However, flow-based warping can create artifacts when optical flow is not accurate, while temporal convolution may suffer from spatial misalignment. We propose ‘Progressive Temporal Feature Alignment Network’, which progressively enriches features extracted from the current frame with the feature warped from neighbouring frames using optical flow. Our approach corrects the spatial misalignment in the temporal feature propagation stage, greatly improving visual quality and temporal consistency of the inpainted videos. Using the proposed architecture, we achieve state-of-the-art performance on the DAVIS and FVI datasets compared to existing deep learning approaches. Code is available at https://github.com/MaureenZOU/TSAM.

We propose a new approach for high resolution semantic image synthesis. It consists of one base image generator and multiple class-specific generators. The base generator generates high quality images based on a segmentation map. To further improve the quality of different objects, we create a bank of Generative Adversarial Networks (GANs) by separately training class-specific models. This has several benefits including – dedicated weights for each class; centrally aligned data for each model; additional training data from other sources, potential of higher resolution and quality; and easy manipulation of a specific object in the scene. Experiments show that our approach can generate high quality images in high resolution while having flexibility of object-level control by using class-specific generators.

We propose YolactEdge, the first competitive instance segmentation approach that runs on small edge devices at real-time speeds. Specifically, YolactEdge runs at up to 30.8 FPS on a Jetson AGX Xavier (and 172.7 FPS on an RTX 2080 Ti) with a ResNet-101 backbone on 550x550 resolution images. To achieve this, we make two improvements to the state-of-the-art image-based real-time method YOLACT [1]: (1) applying TensorRT optimization while carefully trading off speed and accuracy, and (2) a novel feature warping module to exploit temporal redundancy in videos. Experiments on the YouTube VIS and MS COCO datasets demonstrate that YolactEdge produces a 3-5x speed up over existing real-time methods while producing competitive mask and box detection accuracy. We also conduct ablation studies to dissect our design choices and modules. Code and models are available at https://github.com/haotian-liu/yolact_edge.

We propose SinGAN-GIF, an extension of the image-based SinGAN [27] to GIFs or short video snippets. Our method learns the distribution of both the image patches in the GIF as well as their motion patterns. We do so by using a pyramid of 3D and 2D convolutional networks to model temporal information while reducing model parameters and training time, along with an image and a video discriminator. SinGAN-GIF can generate similar looking video samples for natural scenes at different spatial resolutions or temporal frame rates, and can be extended to other video applications like video editing, super resolution, and motion transfer. The project page, with supplementary video results, is: https://rajat95.github.io/singan-gif/.

We present MixNMatch, a conditional generative model that learns to disentangle and encode background, object pose, shape, and texture from real images with minimal supervision, for mix-and-match image generation. We build upon FineGAN, an unconditional generative model, to learn the desired disentanglement and image generator, and leverage adversarial joint image-code distribution matching to learn the latent factor encoders. MixNMatch requires bounding boxes during training to model background, but requires no other supervision. Through extensive experiments, we demonstrate MixNMatch's ability to accurately disentangle, encode, and combine multiple factors for mix-and-match image generation, including sketch2color, cartoon2img, and img2gif applications. Our code/models/demo can be found at https://github.com/Yuheng-Li/MixNMatch

Weakly supervised learning has emerged as a compelling tool for object detection by reducing the need for strong supervision during training. However, major challenges remain: (1) differentiation of object instances can be ambiguous; (2) detectors tend to focus on discriminative parts rather than entire objects; (3) without ground truth, object proposals have to be redundant for high recalls, causing significant memory consumption. Addressing these challenges is difficult, as it often requires to eliminate uncertainties and trivial solutions. To target these issues we develop an instance-aware and context-focused unified framework. It employs an instance-aware self-training algorithm and a learnable Concrete DropBlock while devising a memory-efficient sequential batch back-propagation. Our proposed method achieves state-of-the-art results on COCO (12.1% AP, 24.8% AP50), VOC 2007 (54.9% AP), and VOC 2012 (52.1% AP), improving baselines by great margins. In addition, the proposed method is the first to benchmark ResNet based models and weakly supervised video object detection. Refer to our project page for code, models, and more details: https://github.com/NVlabs/wetectron.

Existing models often leverage co-occurrences between objects and their context to improve recognition accuracy. However, strongly relying on context risks a model's generalizability, especially when typical co-occurrence patterns are absent. This work focuses on addressing such contextual biases to improve the robustness of the learnt feature representations. Our goal is to accurately recognize a category in the absence of its context, without compromising on performance when it co-occurs with context. Our key idea is to decorrelate feature representations of a category from its co-occurring context. We achieve this by learning a feature subspace that explicitly represents categories occurring in the absence of context along side a joint feature subspace that represents both categories and context. Our very simple yet effective method is extensible to two multi-label tasks -- object and attribute classification. On 4 challenging datasets, we demonstrate the effectiveness of our method in reducing contextual bias.

We propose a novel unsupervised generative model that learns to disentangle object identity from other low-level aspects in class-imbalanced data. We first investigate the issues surrounding the assumptions about uniformity made by InfoGAN, and demonstrate its ineffectiveness to properly disentangle object identity in imbalanced data. Our key idea is to make the discovery of the discrete latent factor of variation invariant to identity-preserving transformations in real images, and use that as a signal to learn the appropriate latent distribution representing object identity. Experiments on both artificial (MNIST, 3D cars, 3D chairs, ShapeNet) and real-world (YouTube-Faces) imbalanced datasets demonstrate the effectiveness of our method in disentangling object identity as a latent factor of variation.

We present a method for weakly-supervised action localization based on graph convolutions. In order to find and classify video time segments that correspond to relevant action classes, a system must be able to both identify discriminative time segments in each video, and identify the full extent of each action. Achieving this with weak video level labels requires the system to use similarity and dissimilarity between moments across videos in the training data to understand both how an action appears, as well as the subactions that comprise the action's full extent. However, current methods do not make explicit use of similarity between video moments to inform the localization and classification predictions. We present a novel method that uses graph convolutions to explicitly model similarity between video moments. Our method utilizes similarity graphs that encode appearance and motion, and pushes the state of the art on THUMOS'14, ActivityNet 1.2, and Charades for weakly- supervised action localization.

We present a novel deep neural network architecture for end-to-end scene flow estimation that directly operates on large-scale 3D point clouds. Inspired by Bilateral Convolutional Layers (BCL), we propose novel DownBCL, UpBCL, and CorrBCL operations that restore structural information from unstructured point clouds, and fuse information from two consecutive point clouds. Operating on discrete and sparse permutohedral lattice points, our architectural design is parsimonious in computational cost. Our model can efficiently process a pair of point cloud frames at once with a maximum of 86K points per frame. Our approach achieves state-of-the-art performance on the FlyingThings3D and KITTI Scene Flow 2015 datasets. Moreover, trained on synthetic data, our approach shows great generalization ability on real-world data and on different point densities without fine-tuning.

We propose a novel way of using videos to obtain high precision object proposals for weakly-supervised object detection. Existing weakly-supervised detection approaches use off-the-shelf proposal methods like edge boxes or selective search to obtain candidate boxes. These methods provide high recall but at the expense of thousands of noisy proposals. Thus, the entire burden of finding the few relevant object regions is left to the ensuing object mining step. To mitigate this issue, we focus instead on improving the precision of the initial candidate object proposals. Since we cannot rely on localization annotations, we turn to video and leverage motion cues to automatically estimate the extent of objects to train a Weakly-supervised Region Proposal Network (W-RPN). We use the W-RPN to generate high precision object proposals, which are in turn used to re-rank high recall proposals like edge boxes or selective search according to their spatial overlap. Our W-RPN proposals lead to significant improvement in performance for state-of-the-art weakly-supervised object detection approaches on PASCAL VOC 2007 and 2012.

We propose FineGAN, a novel unsupervised GAN framework, which disentangles the background, object shape, and object appearance to hierarchically generate images of fine-grained object categories. To disentangle the factors without supervision, our key idea is to use information theory to associate each factor to a latent code, and to condition the relationships between the codes in a specific way to induce the desired hierarchy. Through extensive experiments, we show that FineGAN achieves the desired disentanglement to generate realistic and diverse images belonging to fine-grained classes of birds, dogs, and cars. Using FineGAN's automatically learned features, we also cluster real images as a first attempt at solving the novel problem of unsupervised fine-grained object category discovery. Our code/models/demo can be found at https://github.com/kkanshul/finegan.

We present a simple, fully-convolutional model for real-time instance segmentation that achieves 29.8 mAP on MS COCO at 33.5 fps evaluated on a single Titan Xp, which is significantly faster than any previous competitive approach. Moreover, we obtain this result after training on only one GPU. We accomplish this by breaking instance segmentation into two parallel subtasks: (1) generating a set of prototype masks and (2) predicting per-instance mask coefficients. Then we produce instance masks by linearly combining the prototypes with the mask coefficients. We find that because this process doesn't depend on repooling, this approach produces very high-quality masks and exhibits temporal stability for free. Furthermore, we analyze the emergent behavior of our prototypes and show they learn to localize instances on their own in a translation variant manner, despite being fully-convolutional. Finally, we also propose Fast NMS, a drop-in 12 ms faster replacement for standard NMS that only has a marginal performance penalty.

We propose a novel approach that disentangles the identity and pose of objects for image generation. Our model takes as input an ID image and a pose image, and generates an output image with the identity of the ID image and the pose of the pose image. Unlike most previous unsupervised work which rely on cyclic constraints, which can often be brittle, we instead propose to learn this in a self-supervised way. Specifically, we leverage unlabeled videos to automatically construct pseudo ground-truth targets to directly supervise our model. To enforce disentanglement, we propose a novel disentanglement loss, and to improve realism, we propose a pixel-verification loss in which the generated image's pixels must trace back to the ID input. We conduct extensive experiments on both synthetic and real images to demonstrate improved realism, diversity, and ID/pose disentanglement compared to existing methods.

In human learning, it is common to use multiple sources of information jointly. However, most existing feature learning approaches learn from only a single task. In this paper, we propose a novel multi-task deep network to learn generalizable high-level visual representations. Since multitask learning requires annotations for multiple properties of the same training instance, we look to synthetic images to train our network. To overcome the domain difference between real and synthetic data, we employ an unsupervised feature space domain adaptation method based on adversarial learning. Given an input synthetic RGB image, our network simultaneously predicts its surface normal, depth, and instance contour, while also minimizing the feature space domain differences between real and synthetic data. Through extensive experiments, we demonstrate that our network learns more transferable representations compared to single-task baselines. Our learned representation produces state-of-the-art transfer learning results on PASCAL VOC 2007 classification and 2012 detection.

We introduce a novel multimodal machine translation model that utilizes parallel visual and textual information.Our model jointly optimizes the learning of a shared visuallanguage embedding and a translator.The model leverages a visual attention grounding mechanism that links the visual semantics with the corresponding textual semantics.Our approach achieves competitive state-of-the-art results on the Multi30K and the Ambiguous COCO datasets.We also collected a new multilingual multimodal product description dataset to simulate a real-world international online shopping scenario.On this dataset, our visual attention grounding model outperforms other methods by a large margin.

Content popularity prediction has been extensively studied due to its importance and interest for both users and hosts of social media sites like Facebook, Instagram, Twitter, and Pinterest. However, existing work mainly focuses on modeling popularity using a single metric such as the total number of likes or shares. In this work, we propose Diffusion-LSTM, a memory-based deep recurrent network that learns to recursively predict the entire diffusion path of an image through a social network. By combining user social features and image features, and encoding the diffusion path taken thus far with an explicit memory cell, our model predicts the diffusion path of an image more accurately compared to alternate baselines that either encode only image or social features, or lack memory. By mapping individual users to user prototypes, our model can generalize to new users not seen during training. Finally, we demonstrate our model»s capability of generating diffusion trees, and show that the generated trees closely resemble ground-truth trees.

We consider scenarios in which we wish to perform joint scene understanding, object tracking, activity recognition, and other tasks in scenarios in which multiple people are wearing body-worn cameras while a third-person static camera also captures the scene. To do this, we need to establish person-level correspondences across first-and third-person videos, which is challenging because the camera wearer is not visible from his/her own egocentric video, preventing the use of direct feature matching. In this paper, we propose a new semi-Siamese Convolutional Neural Network architecture to address this novel challenge. We formulate the problem as learning a joint embedding space for first-and third-person videos that considers both spatial-and motion-domain cues. A new triplet loss function is designed to minimize the distance between correct first-and third-person matches while maximizing the distance between incorrect ones. This end-to-end approach performs significantly better than several baselines, in part by learning the first-and third-person features optimized for matching jointly with the distance measure itself.

We present a method for localizing facial keypoints on animals by transferring knowledge gained from human faces. Instead of directly finetuning a network trained to detect keypoints on human faces to animal faces (which is sub-optimal since human and animal faces can look quite different), we propose to first adapt the animal images to the pre-trained human detection network by correcting for the differences in animal and human face shape. We first find the nearest human neighbors for each animal image using an unsupervised shape matching method. We use these matches to train a thin plate spline warping network to warp each animal face to look more human-like. The warping network is then jointly finetuned with a pre-trained human facial keypoint detection network using an animal dataset. We demonstrate state-of-the-art results on both horse and sheep facial keypoint detection, and significant improvement over simple finetuning, especially when training data is scarce. Additionally, we present a new dataset with 3717 images with horse face and facial keypoint annotations.

We propose a weakly-supervised approach that takes image-sentence pairs as input and learns to visually ground (i.e., localize) arbitrary linguistic phrases, in the form of spatial attention masks. Specifically, the model is trained with images and their associated image-level captions, without any explicit region-to-phrase correspondence annotations. To this end, we introduce an end-to-end model which learns visual groundings of phrases with two types of carefully designed loss functions. In addition to the standard discriminative loss, which enforces that attended image regions and phrases are consistently encoded, we propose a novel structural loss which makes use of the parse tree structures induced by the sentences. In particular, we ensure complementarity among the attention masks that correspond to sibling noun phrases, and compositionality of attention masks among the children and parent phrases, as defined by the sentence parse tree. We validate the effectiveness of our approach on the Microsoft COCO and Visual Genome datasets.

We propose `Hide-and-Seek', a weakly-supervised framework that aims to improve object localization in images and action localization in videos. Most existing weakly-supervised methods localize only the most discriminative parts of an object rather than all relevant parts, which leads to suboptimal performance. Our key idea is to hide patches in a training image randomly, forcing the network to seek other relevant parts when the most discriminative part is hidden. Our approach only needs to modify the input image and can work with any network designed for object localization. During testing, we do not need to hide any patches. Our Hide-and-Seek approach obtains superior performance compared to previous methods for weakly-supervised object localization on the ILSVRC dataset. We also demonstrate that our framework can be easily extended to weakly-supervised action localization.

In neuroscience, understanding animal behaviors is key to studying their memory patterns. Meanwhile, this is also the most time-consuming and difficult process because it relies heavily on humans to manually annotate the videos recording the animals. In this paper, we present a visual recognition system to automatically annotate animal behaviors to save human annotation costs. By treating the annotation task as a per-frame action classification problem, we can fine-tune a powerful pre-trained convolutional neural network (CNN) for this task. Through extensive experiments, we demonstrate our model not only provides more accurate annotations than alternate automatic methods, but also provides reliable annotations that can replace human annotations for neuroscience experiments.

To achieve growth in the user base of online social networks--(OSN) based applications, word-of-mouth diffusion mechanisms, such as user-to-user invitations, are widely used. This article characterizes the adoption and cascading process of OSN-based applications that grow via user invitations. We analyze a detailed large-scale dataset of a popular Facebook gifting application, iHeart, that contains more than 2 billion entries of user activities generated by 190 million users during a span of 64 weeks. We investigate (1) how users invite their friends to an OSN-based application, (2) how application adoption of an individual user can be predicted, (3) what factors drive the cascading process of application adoptions, and (4) what are the good predictors of the ultimate cascade sizes. We find that sending or receiving a large number of invitations does not necessarily help to recruit new users to iHeart. We also find that the average success ratio of inviters is the most important feature in predicting an adoption of an individual user, which indicates that the effectiveness of inviters has strong predictive power with respect to application adoption. Based on the lessons learned from our analyses, we build and evaluate learning-based models to predict whether a user will adopt iHeart. Our proposed model that utilizes additional activity information of individual users from other similar types of gifting applications can achieve high precision (83%) in predicting adoptions in the target application (i.e., iHeart). We next identify a set of distinctive features that are good predictors of the growth of the application adoptions in terms of final population size. We finally propose a prediction model to infer whether a cascade of application adoption will continue to grow in the future based on observing the initial adoption process. Results show that our proposed model can achieve high precision (over 80%) in predicting large cascades of application adoptions. We believe our work can give an important implication in resource allocation of OSN-based product stakeholders, for example, via targeted marketing.

The status quo approach to training object detectors requires expensive bounding box annotations. Our framework takes a markedly different direction: we transfer tracked object boxes from weakly-labeled videos to weakly-labeled images to automatically generate pseudo ground-truth boxes, which replace manually annotated bounding boxes. We first mine discriminative regions in the weakly-labeled image collection that frequently/rarely appear in the positive/ negative images. We then match those regions to videos and retrieve the corresponding tracked object boxes. Finally, we design a hough transform algorithm to vote for the best box to serve as the pseudo GT for each image, and use them to train an object detector. Together, these lead to state-of-the-art weakly-supervised detection results on the PASCAL 2007 and 2010 datasets.

We present an unsupervised approach that generates a diverse, ranked set of bounding box and segmentation video object proposals-spatio-temporal tubes that localize the foreground objects-in an unannotated video. In contrast to previous unsupervised methods that either track regions initialized in an arbitrary frame or train a fixed model over a cluster of regions, we instead discover a set of easy-togroup instances of an object and then iteratively update its appearance model to gradually detect harder instances in temporally-adjacent frames. Our method first generates a set of spatio-temporal bounding box proposals, and then refines them to obtain pixel-wise segmentation proposals. We demonstrate state-of-the-art segmentation results on the SegTrack v2 dataset, and bounding box tracking results that perform competitively to state-of-the-art supervised tracking methods.

Given a set of poorly aligned images of the same visual concept without any annotations, we propose an algorithm to jointly bring them into pixel-wise correspondence by estimating a FlowWeb representation of the image set. FlowWeb is a fully-connected correspondence flow graph with each node representing an image, and each edge representing the correspondence flow field between a pair of images, i.e. a vector field indicating how each pixel in one image can find a corresponding pixel in the other image. Correspondence flow is related to optical flow but allows for correspondences between visually dissimilar regions if there is evidence they correspond transitively on the graph. Our algorithm starts by initializing all edges of this complete graph with an off-the-shelf, pairwise flow method. We then iteratively update the graph to force it to be more self-consistent. Once the algorithm converges, dense, globally-consistent correspondences can be read off the graph. Our results suggest that FlowWeb improves alignment accuracy over previous pairwise as well as joint alignment methods.

We present a weakly-supervised approach that discovers the spatial extent of relative attributes, given only pairs of ordered images. In contrast to traditional approaches that use global appearance features or rely on keypoint detectors, our goal is to automatically discover the image regions that are relevant to the attribute, even when the attribute's appearance changes drastically across its attribute spectrum. To accomplish this, we first develop a novel formulation that combines a detector with local smoothness to discover a set of coherent visual chains across the image collection. We then introduce an efficient way to generate additional chains anchored on the initial discovered ones. Finally, we automatically identify the most relevant visual chains, and create an ensemble image representation to model the attribute. Through extensive experiments, we demonstrate our method's promise relative to several baselines in modeling relative attributes.

This paper proposes an interactive framework that allows a user to rapidly explore and visualize a large image collection using the medium ofaverage images. Average images have been gaining popularity as means of artistic expression and data visualization, but the creation of compelling examples is a surprisingly laborious and manual process. Our interactive, real-time system provides a way to summarize large amounts of visual data byweightedaverage(s) of an image collection, with the weights reflecting user-indicated importance. The aim is to capture not just the mean of the distribution, but a set of modes discovered via interactive exploration. We pose this exploration in terms of a user interactively "editing" the average image using various types of strokes, brushes and warps, similar to a normal image editor, with each user interaction providing a new constraint to update the average. New weighted averages can be spawned and edited either individually or jointly. Together, these tools allow the user to simultaneously perform two fundamental operations on visual data: user-guided clustering and user-guided alignment, within the same framework. We show that our system is useful for various computer vision and graphics applications.

We present a weakly-supervised visual data mining approach that discovers connections between recurring mid-level visual elements in historic (temporal) and geographic (spatial) image collections, and attempts to capture the underlying visual style. In contrast to existing discovery methods that mine for patterns that remain visually consistent throughout the dataset, our goal is to discover visual elements whose appearance changes due to change in time or location; i.e., exhibit consistent stylistic variations across the label space (date or geo-location). To discover these elements, we first identify groups of patches that are style-sensitive. We then incrementally build correspondences to find the same element across the entire dataset. Finally, we train style-aware regressors that model each element's range of stylistic differences. We apply our approach to date and geo-location prediction and show substantial improvement over several baselines that do not model visual style. We also demonstrate the method's effectiveness on the related task of fine-grained classification.

We present a video summarization approach for egocentric or “wearable” camera data. Given hours of video, the proposed method produces a compact storyboard summary of the camera wearer's day. In contrast to traditional keyframe selection techniques, the resulting summary focuses on the most important objects and people with which the camera wearer interacts. To accomplish this, we develop region cues indicative of high-level saliency in egocentric video — such as the nearness to hands, gaze, and frequency of occurrence — and learn a regressor to predict the relative importance of any new region based on these cues. Using these predictions and a simple form of temporal event detection, our method selects frames for the storyboard that reflect the key object-driven happenings. Critically, the approach is neither camera-wearer-specific nor object-specific; that means the learned importance metric need not be trained for a given user or context, and it can predict the importance of objects and people that have never been seen previously. Our results with 17 hours of egocentric data show the method's promise relative to existing techniques for saliency and summarization.

How can knowing about some categories help us to discover new ones in unlabeled images? Unsupervised visual category discovery is useful to mine for recurring objects without human supervision, but existing methods assume no prior information and thus tend to perform poorly for cluttered scenes with multiple objects. We propose to leverage knowledge about previously learned categories to enable more accurate discovery, and address challenges in estimating their familiarity in unsegmented, unlabeled images. We introduce two variants of a novel object-graph descriptor to encode the 2D and 3D spatial layout of object-level co-occurrence patterns relative to an unfamiliar region and show that by using them to model the interaction between an image’s known and unknown objects, we can better detect new visual categories. Rather than mine for all categories from scratch, our method identifies new objects while drawing on useful cues from familiar ones. We evaluate our approach on several benchmark data sets and demonstrate clear improvements in discovery over conventional purely appearance-based baselines.

We present an approach to discover novel faces in untagged photo collections by leveraging the “social context” of co-occurring people. Our idea exploits the social nature of consumer photos, in which people of the same clique (family, team, class, friends) often appear together. Initially, the system trains detector s for any individuals with tagged instances in the collection. Then, for each untagged image, it isolates any unfamiliar faces. Among those, it discovers novel face clusters by leveraging both their appearance, as well as descriptors encoding the (predicted) familiar fa ces with which the unfamiliar faces co-occur. The resulting discovered people can then be presented to a user for nametagging, thereby efficiently propagating manually provide d labels. Our experiments with real consumer photo collections demonstrate that the system outperforms baseline approaches that either lack any social context model, or rely solely on the appearance of co-occurring faces. Furthermore, we show it can successfully use the discovered models it forms to auto-tag unseen faces in a new collection.

Objects vary in their visual complexity, yet existing discovery methods perform “batch” clustering, paying equal attention to all instances simultaneously — regardless of the strength of their appearance or context cues. We propose a self-paced approach that instead focuses on the easiest instances first, and progressively expands its repertoire to include more complex objects. Easier regions are defined as those with both high likelihood of generic objectness and high familiarity of surrounding objects. At each cycle of the discovery process, we re-estimate the easiness of each subwindow in the pool of unlabeled images, and then retrieve a single prominent cluster from among the easiest instances. Critically, as the system gradually accumulates models, each new (more difficult) discovery benefits from the context provided by earlier discoveries. Our experiments demonstrate the clear advantages of self-paced discovery relative to conventional batch approaches, including both more accurate summarization as well as stronger predictive models for novel data.

We present an approach to discover and segment foreground object(s) in video. Given an unannotated video sequence, the method first identifies object-like regions in any frame according to both static and dynamic cues. We then compute a series of binary partitions among those candidate “key-segments” to discover hypothesis groups with persistent appearance and motion. Finally, using each ranked hypothesis in turn, we estimate a pixel-level object labeling across all frames, where (a) the foreground likelihood depends on both the hypothesis's appearance as well as a novel localization prior based on partial shape matching, and (b) the background likelihood depends on cues pulled from the key-segments' (possibly diverse) surroundings observed across the sequence. Compared to existing methods, our approach automatically focuses on the persistent foreground regions of interest while resisting oversegmentation. We apply our method to challenging benchmark videos, and show competitive or better results than the state-of-the-art.

We present ShadowDraw, a system for guiding the freeform drawing of objects. As the user draws, ShadowDraw dynamically updates a shadow image underlying the user's strokes. The shadows are suggestive of object contours that guide the user as they continue drawing. This paradigm is similar to tracing, with two major differences. First, we do not provide a single image from which the user can trace; rather ShadowDraw automatically blends relevant images from a large database to construct the shadows. Second, the system dynamically adapts to the user's drawings in real-time and produces suggestions accordingly. ShadowDraw works by efficiently matching local edge patches between the query, constructed from the current drawing, and a database of images. A hashing technique enforces both local and global similarity and provides sufficient speed for interactive feedback. Shadows are created by aggregating the edge maps from the best database matches, spatially weighted by their match scores. We test our approach with human subjects and show comparisons between the drawings that were produced with and without the system. The results show that our system produces more realistically proportioned line drawings.

How can knowing about some categories help us to discover new ones in unlabeled images? Unsupervised visual category discovery is useful to mine for recurring objects without human supervision, but existing methods assume no prior information and thus tend to perform poorly for cluttered scenes with multiple objects. We propose to leverage knowledge about previously learned categories to enable more accurate discovery. We introduce a novel object-graph descriptor to encode the layout of object-level co-occurrence patterns relative to an unfamiliar region, and show that by using it to model the interaction between an image's known and unknown objects we can better detect new visual categories. Rather than mine for all categories from scratch, our method identifies new objects while drawing on useful cues from familiar ones. We evaluate our approach on benchmark datasets and demonstrate clear improvements in discovery over conventional purely appearance-based baselines.

We present a method to segment a collection of unlabeled images while exploiting automatically discovered appearance patterns shared between them. Given an unlabeled pool of multi-object images, we first detect any visual clusters present among their sub-regions, where inter-region similarity is measured according to both appearance and contextual layout. Then, using each initial segment as a seed, we solve a graph cuts problem to refine its boundary — enforcing preferences to include nearby regions that agree with an ensemble of representative regions discovered for that cluster, and exclude those regions that resemble familiar objects. Through extensive experiments, we show that the segmentations computed jointly on the collection agree more closely with true object boundaries, when compared to either a bottom-up baseline or a graph cuts foreground segmentation that can only access cues from a single image.

Can we discover common object shapes within unlabeled multi-category collections of images? While often a critical cue at the category-level, contour matches can be difficult to isolate reliably from edge clutter—even within labeled images from a known class, let alone unlabeled examples. We propose a shape discovery method in which local appearance (patch) matches serve to anchor the surrounding edge fragments, yielding a more reliable affinity function for images that accounts for both shape and appearance. Spectral clustering from the initial affinities provides candidate object clusters. Then, we compute the within-cluster match patterns to discern foreground edges from clutter, attributing higher weight to edges more likely to belong to a common object. In addition to discovering the object contours in each image, we show how to summarize what is found with prototypical shapes. Our results on benchmark datasets demonstrate the approach can successfully discover shapes from unlabeled images.

We present a method to automatically discover meaningful features in unlabeled image collections. Each image is decomposed into semi-local features that describe neighborhood appearance and geometry. The goal is to determine for each image which of these parts are most relevant, given the image content in the remainder of the collection. Our method first computes an initial image-level grouping based on feature correspondences, and then iteratively refines cluster assignments based on the evolving intra-cluster pattern of local matches. As a result, the significance attributed to each feature influences an image’s cluster membership, while related images in a cluster affect the estimated significance of their features. We show that this mutual reinforcement of object-level and feature-level similarity improves unsupervised image clustering, and apply the technique to automatically discover categories and foreground regions in images from benchmark datasets. 1