Number it: Temporal Grounding Videos like Flipping Manga

Video Large Language Models (Vid-LLMs) have made remarkable advancements in comprehending video content for QA dialogue. However, they struggle to extend this visual understanding to tasks requiring precise temporal localization, known as Video Temporal Grounding (VTG). To address this gap, we introduce Number-Prompt (NumPro), a novel method that empowers Vid-LLMs to bridge visual comprehension with temporal grounding by adding unique numerical identifiers to each video frame. Treating a video as a sequence of numbered frame images, NumPro transforms VTG into an intuitive process: flipping through manga panels in sequence. This allows Vid-LLMs to "read" event timelines, accurately linking visual content with corresponding temporal information. Our experiments demonstrate that NumPro significantly boosts VTG performance of top-tier Vid-LLMs without additional computational cost. Furthermore, fine-tuning on a NumPro-enhanced dataset defines a new state-of-the-art for VTG, surpassing previous top-performing methods by up to 6.9\% in mIoU for moment retrieval and 8.5\% in mAP for highlight detection. The code will be available at https://github.com/yongliang-wu/NumPro.

MMAR: Towards Lossless Multi-Modal Auto-Regressive Probabilistic Modeling

Recent advancements in multi-modal large language models have propelled the development of joint probabilistic models capable of both image understanding and generation. However, we have identified that recent methods inevitably suffer from loss of image information during understanding task, due to either image discretization or diffusion denoising steps. To address this issue, we propose a novel Multi-Modal Auto-Regressive (MMAR) probabilistic modeling framework. Unlike discretization line of method, MMAR takes in continuous-valued image tokens to avoid information loss. Differing from diffusion-based approaches, we disentangle the diffusion process from auto-regressive backbone model by employing a light-weight diffusion head on top each auto-regressed image patch embedding. In this way, when the model transits from image generation to understanding through text generation, the backbone model's hidden representation of the image is not limited to the last denoising step. To successfully train our method, we also propose a theoretically proven technique that addresses the numerical stability issue and a training strategy that balances the generation and understanding task goals. Through extensive evaluations on 18 image understanding benchmarks, MMAR demonstrates much more superior performance than other joint multi-modal models, matching the method that employs pretrained CLIP vision encoder, meanwhile being able to generate high quality images at the same time. We also showed that our method is scalable with larger data and model size.

EE-MLLM: A Data-Efficient and Compute-Efficient Multimodal Large Language Model

In the realm of multimodal research, numerous studies leverage substantial image-text pairs to conduct modal alignment learning, transforming Large Language Models (LLMs) into Multimodal LLMs and excelling in a variety of visual-language tasks. The prevailing methodologies primarily fall into two categories: self-attention-based and cross-attention-based methods. While self-attention-based methods offer superior data efficiency due to their simple MLP architecture, they often suffer from lower computational efficiency due to concatenating visual and textual tokens as input for LLM. Conversely, cross-attention-based methods, although less data-efficient due to additional learnable parameters, exhibit higher computational efficiency by avoiding long sequence input for LLM. To address these trade-offs, we introduce the Data-Efficient and Compute-Efficient Multimodal Large Language Model (EE-MLLM). Without introducing additional modules or learnable parameters, EE-MLLM achieves both data and compute efficiency. Specifically, we modify the original self-attention mechanism in MLLM to a composite attention mechanism. This mechanism has two key characteristics: 1) Eliminating the computational overhead of self-attention within visual tokens to achieve compute efficiency, and 2) Reusing the weights on each layer of LLM to facilitate effective modality alignment between vision and language for data efficiency. Experimental results demonstrate the effectiveness of EE-MLLM across a range of benchmarks, including general-purpose datasets like MMBench and SeedBench, as well as fine-grained tasks such as TextVQA and DocVQA.

Visual Perception by Large Language Model's Weights

Existing Multimodal Large Language Models (MLLMs) follow the paradigm that perceives visual information by aligning visual features with the input space of Large Language Models (LLMs), and concatenating visual tokens with text tokens to form a unified sequence input for LLMs. These methods demonstrate promising results on various vision-language tasks but are limited by the high computational effort due to the extended input sequence resulting from the involvement of visual tokens. In this paper, instead of input space alignment, we propose a novel parameter space alignment paradigm that represents visual information as model weights. For each input image, we use a vision encoder to extract visual features, convert features into perceptual weights, and merge the perceptual weights with LLM's weights. In this way, the input of LLM does not require visual tokens, which reduces the length of the input sequence and greatly improves efficiency. Following this paradigm, we propose VLoRA with the perceptual weights generator. The perceptual weights generator is designed to convert visual features to perceptual weights with low-rank property, exhibiting a form similar to LoRA. The experimental results show that our VLoRA achieves comparable performance on various benchmarks for MLLMs, while significantly reducing the computational costs for both training and inference. The code and models will be made open-source.

Multi-Modal Generative Embedding Model

Most multi-modal tasks can be formulated into problems of either generation or embedding. Existing models usually tackle these two types of problems by decoupling language modules into a text decoder for generation, and a text encoder for embedding. To explore the minimalism of multi-modal paradigms, we attempt to achieve only one model per modality in this work. We propose a Multi-Modal Generative Embedding Model (MM-GEM), whereby the generative and embedding objectives are encapsulated in one Large Language Model. We also propose a PoolAggregator to boost efficiency and enable the ability of fine-grained embedding and generation. A surprising finding is that these two objectives do not significantly conflict with each other. For example, MM-GEM instantiated from ViT-Large and TinyLlama shows competitive performance on benchmarks for multimodal embedding models such as cross-modal retrieval and zero-shot classification, while has good ability of image captioning. Additionally, MM-GEM can seamlessly execute region-level image caption generation and retrieval tasks. Besides, the advanced text model in MM-GEM brings over 5% improvement in Recall@1 for long text and image retrieval.

ReGenNet: Towards Human Action-Reaction Synthesis

Humans constantly interact with their surrounding environments. Current human-centric generative models mainly focus on synthesizing humans plausibly interacting with static scenes and objects, while the dynamic human action-reaction synthesis for ubiquitous causal human-human interactions is less explored. Human-human interactions can be regarded as asymmetric with actors and reactors in atomic interaction periods. In this paper, we comprehensively analyze the asymmetric, dynamic, synchronous, and detailed nature of human-human interactions and propose the first multi-setting human action-reaction synthesis benchmark to generate human reactions conditioned on given human actions. To begin with, we propose to annotate the actor-reactor order of the interaction sequences for the NTU120, InterHuman, and Chi3D datasets. Based on them, a diffusion-based generative model with a Transformer decoder architecture called ReGenNet together with an explicit distance-based interaction loss is proposed to predict human reactions in an online manner, where the future states of actors are unavailable to reactors. Quantitative and qualitative results show that our method can generate instant and plausible human reactions compared to the baselines, and can generalize to unseen actor motions and viewpoint changes.

Spatial-Semantic Collaborative Cropping for User Generated Content

A large amount of User Generated Content (UGC) is uploaded to the Internet daily and displayed to people world-widely through the client side (e.g., mobile and PC). This requires the cropping algorithms to produce the aesthetic thumbnail within a specific aspect ratio on different devices. However, existing image cropping works mainly focus on landmark or landscape images, which fail to model the relations among the multi-objects with the complex background in UGC. Besides, previous methods merely consider the aesthetics of the cropped images while ignoring the content integrity, which is crucial for UGC cropping. In this paper, we propose a Spatial-Semantic Collaborative cropping network (S2CNet) for arbitrary user generated content accompanied by a new cropping benchmark. Specifically, we first mine the visual genes of the potential objects. Then, the suggested adaptive attention graph recasts this task as a procedure of information association over visual nodes. The underlying spatial and semantic relations are ultimately centralized to the crop candidate through differentiable message passing, which helps our network efficiently to preserve both the aesthetics and the content integrity. Extensive experiments on the proposed UGCrop5K and other public datasets demonstrate the superiority of our approach over state-of-the-art counterparts. Our project is available at https://github.com/suyukun666/S2CNet.

Inter-X: Towards Versatile Human-Human Interaction Analysis

The analysis of the ubiquitous human-human interactions is pivotal for understanding humans as social beings. Existing human-human interaction datasets typically suffer from inaccurate body motions, lack of hand gestures and fine-grained textual descriptions. To better perceive and generate human-human interactions, we propose Inter-X, a currently largest human-human interaction dataset with accurate body movements and diverse interaction patterns, together with detailed hand gestures. The dataset includes ~11K interaction sequences and more than 8.1M frames. We also equip Inter-X with versatile annotations of more than 34K fine-grained human part-level textual descriptions, semantic interaction categories, interaction order, and the relationship and personality of the subjects. Based on the elaborate annotations, we propose a unified benchmark composed of 4 categories of downstream tasks from both the perceptual and generative directions. Extensive experiments and comprehensive analysis show that Inter-X serves as a testbed for promoting the development of versatile human-human interaction analysis. Our dataset and benchmark will be publicly available for research purposes.

Text-Only Image Captioning with Multi-Context Data Generation

Text-only Image Captioning (TIC) is an approach that aims to construct a model solely based on text that can accurately describe images. Recently, diffusion models have demonstrated remarkable capabilities in generating high-quality images that are semantically coherent with given texts. This presents an opportunity to generate synthetic training images for TIC. However, we have identified a challenge that the images generated from simple descriptions typically exhibit a single perspective with one or limited contexts, which is not aligned with the complexity of real-world scenes in the image domain. In this paper, we propose a novel framework that addresses this issue by introducing multi-context data generation. Starting with an initial text corpus, our framework employs a large language model to select multiple sentences that describe the same scene from various perspectives. These sentences are then summarized into a single sentence with multiple contexts. We generate simple images using the straightforward sentences and complex images using the summarized sentences through diffusion models. Finally, we train the model exclusively using the synthetic image-text pairs obtained from this process. Experimental results demonstrate that our proposed framework effectively tackles the central challenge we have identified, achieving the state-of-the-art performance on popular datasets such as MSCOCO, Flickr30k, and SS1M.

A Similarity Alignment Model for Video Copy Segment Matching

With the development of multimedia technology, Video Copy Detection has been a crucial problem for social media platforms. Meta AI hold Video Similarity Challenge on CVPR 2023 to push the technology forward. In this report, we share our winner solutions on Matching Track. We propose a Similarity Alignment Model(SAM) for video copy segment matching. Our SAM exhibits superior performance compared to other competitors, with a 0.108 / 0.144 absolute improvement over the second-place competitor in Phase 1 / Phase 2. Code is available at https://github.com/FeipengMa6/VSC22-Submission/tree/main/VSC22-Matching-Track-1st.