Abstract:Though diffusion-based video generation has witnessed rapid progress, the inference results of existing models still exhibit unsatisfactory temporal consistency and unnatural dynamics. In this paper, we delve deep into the noise initialization of video diffusion models, and discover an implicit training-inference gap that attributes to the unsatisfactory inference quality. Our key findings are: 1) the spatial-temporal frequency distribution of the initial latent at inference is intrinsically different from that for training, and 2) the denoising process is significantly influenced by the low-frequency components of the initial noise. Motivated by these observations, we propose a concise yet effective inference sampling strategy, FreeInit, which significantly improves temporal consistency of videos generated by diffusion models. Through iteratively refining the spatial-temporal low-frequency components of the initial latent during inference, FreeInit is able to compensate the initialization gap between training and inference, thus effectively improving the subject appearance and temporal consistency of generation results. Extensive experiments demonstrate that FreeInit consistently enhances the generation results of various text-to-video generation models without additional training.
Abstract:Video generation has witnessed significant advancements, yet evaluating these models remains a challenge. A comprehensive evaluation benchmark for video generation is indispensable for two reasons: 1) Existing metrics do not fully align with human perceptions; 2) An ideal evaluation system should provide insights to inform future developments of video generation. To this end, we present VBench, a comprehensive benchmark suite that dissects "video generation quality" into specific, hierarchical, and disentangled dimensions, each with tailored prompts and evaluation methods. VBench has three appealing properties: 1) Comprehensive Dimensions: VBench comprises 16 dimensions in video generation (e.g., subject identity inconsistency, motion smoothness, temporal flickering, and spatial relationship, etc). The evaluation metrics with fine-grained levels reveal individual models' strengths and weaknesses. 2) Human Alignment: We also provide a dataset of human preference annotations to validate our benchmarks' alignment with human perception, for each evaluation dimension respectively. 3) Valuable Insights: We look into current models' ability across various evaluation dimensions, and various content types. We also investigate the gaps between video and image generation models. We will open-source VBench, including all prompts, evaluation methods, generated videos, and human preference annotations, and also include more video generation models in VBench to drive forward the field of video generation.
Abstract:This work aims to learn a high-quality text-to-video (T2V) generative model by leveraging a pre-trained text-to-image (T2I) model as a basis. It is a highly desirable yet challenging task to simultaneously a) accomplish the synthesis of visually realistic and temporally coherent videos while b) preserving the strong creative generation nature of the pre-trained T2I model. To this end, we propose LaVie, an integrated video generation framework that operates on cascaded video latent diffusion models, comprising a base T2V model, a temporal interpolation model, and a video super-resolution model. Our key insights are two-fold: 1) We reveal that the incorporation of simple temporal self-attentions, coupled with rotary positional encoding, adequately captures the temporal correlations inherent in video data. 2) Additionally, we validate that the process of joint image-video fine-tuning plays a pivotal role in producing high-quality and creative outcomes. To enhance the performance of LaVie, we contribute a comprehensive and diverse video dataset named Vimeo25M, consisting of 25 million text-video pairs that prioritize quality, diversity, and aesthetic appeal. Extensive experiments demonstrate that LaVie achieves state-of-the-art performance both quantitatively and qualitatively. Furthermore, we showcase the versatility of pre-trained LaVie models in various long video generation and personalized video synthesis applications.
Abstract:In this paper, we present the details of Women in Computer Vision Workshop - WiCV 2023, organized alongside the hybrid CVPR 2023 in Vancouver, Canada. WiCV aims to amplify the voices of underrepresented women in the computer vision community, fostering increased visibility in both academia and industry. We believe that such events play a vital role in addressing gender imbalances within the field. The annual WiCV@CVPR workshop offers a) opportunity for collaboration between researchers from minority groups, b) mentorship for female junior researchers, c) financial support to presenters to alleviate finanacial burdens and d) a diverse array of role models who can inspire younger researchers at the outset of their careers. In this paper, we present a comprehensive report on the workshop program, historical trends from the past WiCV@CVPR events, and a summary of statistics related to presenters, attendees, and sponsorship for the WiCV 2023 workshop.
Abstract:In this paper, we uncover the untapped potential of diffusion U-Net, which serves as a "free lunch" that substantially improves the generation quality on the fly. We initially investigate the key contributions of the U-Net architecture to the denoising process and identify that its main backbone primarily contributes to denoising, whereas its skip connections mainly introduce high-frequency features into the decoder module, causing the network to overlook the backbone semantics. Capitalizing on this discovery, we propose a simple yet effective method-termed "FreeU" - that enhances generation quality without additional training or finetuning. Our key insight is to strategically re-weight the contributions sourced from the U-Net's skip connections and backbone feature maps, to leverage the strengths of both components of the U-Net architecture. Promising results on image and video generation tasks demonstrate that our FreeU can be readily integrated to existing diffusion models, e.g., Stable Diffusion, DreamBooth, ModelScope, Rerender and ReVersion, to improve the generation quality with only a few lines of code. All you need is to adjust two scaling factors during inference. Project page: https://chenyangsi.top/FreeU/.
Abstract:Recently there has been a series of studies in knowledge graph embedding (KGE), which attempts to learn the embeddings of the entities and relations as numerical vectors and mathematical mappings via machine learning (ML). However, there has been limited research that applies KGE for industrial problems in manufacturing. This paper investigates whether and to what extent KGE can be used for an important problem: quality monitoring for welding in manufacturing industry, which is an impactful process accounting for production of millions of cars annually. The work is in line with Bosch research of data-driven solutions that intends to replace the traditional way of destroying cars, which is extremely costly and produces waste. The paper tackles two very challenging questions simultaneously: how large the welding spot diameter is; and to which car body the welded spot belongs to. The problem setting is difficult for traditional ML because there exist a high number of car bodies that should be assigned as class labels. We formulate the problem as link prediction, and experimented popular KGE methods on real industry data, with consideration of literals. Our results reveal both limitations and promising aspects of adapted KGE methods.
Abstract:Diffusion models arise as a powerful generative tool recently. Despite the great progress, existing diffusion models mainly focus on uni-modal control, i.e., the diffusion process is driven by only one modality of condition. To further unleash the users' creativity, it is desirable for the model to be controllable by multiple modalities simultaneously, e.g., generating and editing faces by describing the age (text-driven) while drawing the face shape (mask-driven). In this work, we present Collaborative Diffusion, where pre-trained uni-modal diffusion models collaborate to achieve multi-modal face generation and editing without re-training. Our key insight is that diffusion models driven by different modalities are inherently complementary regarding the latent denoising steps, where bilateral connections can be established upon. Specifically, we propose dynamic diffuser, a meta-network that adaptively hallucinates multi-modal denoising steps by predicting the spatial-temporal influence functions for each pre-trained uni-modal model. Collaborative Diffusion not only collaborates generation capabilities from uni-modal diffusion models, but also integrates multiple uni-modal manipulations to perform multi-modal editing. Extensive qualitative and quantitative experiments demonstrate the superiority of our framework in both image quality and condition consistency.
Abstract:Diffusion models gain increasing popularity for their generative capabilities. Recently, there have been surging needs to generate customized images by inverting diffusion models from exemplar images. However, existing inversion methods mainly focus on capturing object appearances. How to invert object relations, another important pillar in the visual world, remains unexplored. In this work, we propose ReVersion for the Relation Inversion task, which aims to learn a specific relation (represented as "relation prompt") from exemplar images. Specifically, we learn a relation prompt from a frozen pre-trained text-to-image diffusion model. The learned relation prompt can then be applied to generate relation-specific images with new objects, backgrounds, and styles. Our key insight is the "preposition prior" - real-world relation prompts can be sparsely activated upon a set of basis prepositional words. Specifically, we propose a novel relation-steering contrastive learning scheme to impose two critical properties of the relation prompt: 1) The relation prompt should capture the interaction between objects, enforced by the preposition prior. 2) The relation prompt should be disentangled away from object appearances. We further devise relation-focal importance sampling to emphasize high-level interactions over low-level appearances (e.g., texture, color). To comprehensively evaluate this new task, we contribute ReVersion Benchmark, which provides various exemplar images with diverse relations. Extensive experiments validate the superiority of our approach over existing methods across a wide range of visual relations.
Abstract:Contrast-enhanced T1 (T1ce) is one of the most essential magnetic resonance imaging (MRI) modalities for diagnosing and analyzing brain tumors, especially gliomas. In clinical practice, common MRI modalities such as T1, T2, and fluid attenuation inversion recovery are relatively easy to access while T1ce is more challenging considering the additional cost and potential risk of allergies to the contrast agent. Therefore, it is of great clinical necessity to develop a method to synthesize T1ce from other common modalities. Current paired image translation methods typically have the issue of requiring a large amount of paired data and do not focus on specific regions of interest, e.g., the tumor region, in the synthesization process. To address these issues, we propose a Difficulty-perceived common-to-T1ce Semi-Supervised multimodal MRI Synthesis network (DS3-Net), involving both paired and unpaired data together with dual-level knowledge distillation. DS3-Net predicts a difficulty map to progressively promote the synthesis task. Specifically, a pixelwise constraint and a patchwise contrastive constraint are guided by the predicted difficulty map. Through extensive experiments on the publiclyavailable BraTS2020 dataset, DS3-Net outperforms its supervised counterpart in each respect. Furthermore, with only 5% paired data, the proposed DS3-Net achieves competitive performance with state-of-theart image translation methods utilizing 100% paired data, delivering an average SSIM of 0.8947 and an average PSNR of 23.60.
Abstract:Multi-modal magnetic resonance (MR) imaging provides great potential for diagnosing and analyzing brain gliomas. In clinical scenarios, common MR sequences such as T1, T2 and FLAIR can be obtained simultaneously in a single scanning process. However, acquiring contrast enhanced modalities such as T1ce requires additional time, cost, and injection of contrast agent. As such, it is clinically meaningful to develop a method to synthesize unavailable modalities which can also be used as additional inputs to downstream tasks (e.g., brain tumor segmentation) for performance enhancing. In this work, we propose an end-to-end framework named Modality-Level Attention Fusion Network (MAF-Net), wherein we innovatively conduct patchwise contrastive learning for extracting multi-modal latent features and dynamically assigning attention weights to fuse different modalities. Through extensive experiments on BraTS2020, our proposed MAF-Net is found to yield superior T1ce synthesis performance (SSIM of 0.8879 and PSNR of 22.78) and accurate brain tumor segmentation (mean Dice scores of 67.9%, 41.8% and 88.0% on segmenting the tumor core, enhancing tumor and whole tumor).