论文检索

Nannan Li, Kevin J Shih, Bryan A. Plummer

Human pose transfer synthesizes new view(s) of a person for a given pose. Recent work achieves this via self-reconstruction, which disentangles a person's pose and texture information by breaking down the person into several parts, then recombines them to reconstruct the person. However, this part-level disentanglement preserves some pose information that can create unwanted artifacts. In this paper, we propose Pose Transfer by Permuting Textures, a self-driven human pose transfer approach that disentangles pose from texture at the patch-level. Specifically, we remove pose from an input image by permuting image patches so only texture information remains. Then we reconstruct the input image by sampling from the permuted textures to achieve patch-level disentanglement. To reduce the noise and recover clothing shape information from the permuted patches, we employ encoders with multiple kernel sizes in a triple branch network. Extensive experiments on DeepFashion and Market-1501 show that our model improves the quality of generated images in terms of FID, LPIPS and SSIM over other self-driven methods, and even outperforming some fully-supervised methods. A user study also shows that among self-driven approaches, images generated by our method are preferred in 68% of cases over prior work. Code is available at https://github.com/NannanLi999/pt_square.

Cheng-Kun Yang, Min-Hung Chen, Yung-Yu Chuang, Yen-Yu Lin

We present a Multimodal Interlaced Transformer (MIT) that jointly considers 2D and 3D data for weakly supervised point cloud segmentation. Research studies have shown that 2D and 3D features are complementary for point cloud segmentation. However, existing methods require extra 2D annotations to achieve 2D-3D information fusion. Considering the high annotation cost of point clouds, effective 2D and 3D feature fusion based on weakly supervised learning is in great demand. To this end, we propose a transformer model with two encoders and one decoder for weakly supervised point cloud segmentation using only scene-level class tags. Specifically, the two encoders compute the self-attended features for 3D point clouds and 2D multi-view images, respectively. The decoder implements interlaced 2D-3D cross-attention and carries out implicit 2D and 3D feature fusion. We alternately switch the roles of queries and key-value pairs in the decoder layers. It turns out that the 2D and 3D features are iteratively enriched by each other. Experiments show that it performs favorably against existing weakly supervised point cloud segmentation methods by a large margin on the S3DIS and ScanNet benchmarks.

Siddharth Gururani, Arun Mallya, Ting-Chun Wang, Rafael Valle, Ming-Yu Liu

Animating portraits using speech has received growing attention in recent years, with various creative and practical use cases. An ideal generated video should have good lip sync with the audio, natural facial expressions and head motions, and high frame quality. In this work, we present SPACE, which uses speech and a single image to generate high-resolution, and expressive videos with realistic head pose, without requiring a driving video. It uses a multi-stage approach, combining the controllability of facial landmarks with the high-quality synthesis power of a pretrained face generator. SPACE also allows for the control of emotions and their intensities. Our method outperforms prior methods in objective metrics for image quality and facial motions and is strongly preferred by users in pair-wise comparisons. Please visit the project page to view the videos and to see more results: https://research.nvidia.com/labs/dir/space.

Donghwan Seo, Abhijith Punnappurath, Luxi Zhao, Abdelrahman Abdelhamed, Sai Kiran Tedla, Sanguk Park, Jihwan Choe, Michael S. Brown

Computer graphics (CG) rendering platforms produce imagery with ever-increasing photo realism. The narrowing domain gap between real and synthetic imagery makes it possible to use CG images as training data for deep learning models targeting high-level computer vision tasks, such as autonomous driving and semantic segmentation. CG images, however, are currently not suitable for low-level vision tasks targeting RAW sensor images. This is because RAW images are encoded in sensor-specific color spaces and incur pre-white-balance color casts caused by the sensor's response to scene illumination. CG images are rendered directly to a device-independent perceptual color space without needing white balancing. As a result, it is necessary to apply a mapping procedure to close the domain gap between graphics and RAW images. To this end, we introduce a framework to process graphics images to mimic RAW sensor images accurately. Our approach allows a one-to-many mapping, where a single graphics image can be transformed to match multiple sensors and multiple scene illuminations. In addition, our approach requires only a handful of example RAW-DNG files from the target sensor as parameters for the mapping process. We compare our method to alternative strategies and show that our approach produces more realistic RAW images and provides better results on three low-level vision tasks: RAW denoising, illumination estimation, and neural rendering for night photography. Finally, as part of this work, we provide a dataset of 292 realistic CG images for training low-light imaging models.

Yifan Yang, Shuhai Zhang, Zixiong Huang, Yubing Zhang, Mingkui Tan

Neural Radiance Fields (NeRF) is a revolutionary approach for rendering scenes by sampling a single ray per pixel and it has demonstrated impressive capabilities in novel-view synthesis from static scene images. However, in practice, we usually need to recover NeRF from unconstrained image collections, which poses two challenges: 1) the images often have dynamic changes in appearance because of different capturing time and camera settings; 2) the images may contain transient objects such as humans and cars, leading to occlusion and ghosting artifacts. Conventional approaches seek to address these challenges by locally utilizing a single ray to synthesize a color of a pixel. In contrast, humans typically perceive appearance and objects by globally utilizing information across multiple pixels. To mimic the perception process of humans, in this paper, we propose Cross-Ray NeRF (CR-NeRF) that leverages interactive information across multiple rays to synthesize occlusion-free novel views with the same appearances as the images. Specifically, to model varying appearances, we first propose to represent multiple rays with a novel cross-ray feature and then recover the appearance by fusing global statistics, i.e., feature covariance of the rays and the image appearance. Moreover, to avoid occlusion introduced by transient objects, we propose a transient objects handler and introduce a grid sampling strategy for masking out the transient objects. We theoretically find that leveraging correlation across multiple rays promotes capturing more global information. Moreover, extensive experimental results on large real-world datasets verify the effectiveness of CR-NeRF.

Hyundong Jin, Gyeong-hyeon Kim, Chanho Ahn, Eunwoo Kim

Deep neural networks suffer from catastrophic forgetting in continual learning, where they tend to lose information about previously learned tasks when optimizing a new incoming task. Recent strategies isolate the important parameters for previous tasks to retain old knowledge while learning the new task. However, using the fixed old knowledge might act as an obstacle to capturing novel representations. To overcome this limitation, we propose a framework that evolves the previously allocated parameters by absorbing the knowledge of the new task. The approach performs under two different networks. The base network learns knowledge of sequential tasks, and the sparsity-inducing hypernetwork generates parameters for each time step for evolving old knowledge. The generated parameters transform old parameters of the base network to reflect the new knowledge. We design the hypernetwork to generate sparse parameters conditional to the task-specific information and the structural information of the base network. We evaluate the proposed approach on class-incremental and task-incremental learning scenarios for image classification and video action recognition tasks. Experimental results show that the proposed method consistently outperforms a large variety of continual learning approaches for those scenarios by evolving old knowledge.

Kyle Sargent, Jing Yu Koh, Han Zhang, Huiwen Chang, Charles Herrmann, Pratul Srinivasan, Jiajun Wu, Deqing Sun

Recent work has shown the possibility of training generative models of 3D content from 2D image collections on small datasets corresponding to a single object class, such as human faces, animal faces, or cars. However, these models struggle on larger, more complex datasets. To model diverse and unconstrained image collections such as ImageNet, we present VQ3D, which introduces a NeRF-based decoder into a two-stage vector-quantized autoencoder. Our Stage 1 allows for the reconstruction of an input image and the ability to change the camera position around the image, and our Stage 2 allows for the generation of new 3D scenes. VQ3D is capable of generating and reconstructing 3D-aware images from the 1000-class ImageNet dataset of 1.2 million training images, and achieves a competitive ImageNet generation FID score of 16.8.

Shuangrui Ding, Peisen Zhao, Xiaopeng Zhang, Rui Qian, Hongkai Xiong, Qi Tian

Transformers have become the primary backbone of the computer vision community due to their impressive performance. However, the unfriendly computation cost impedes their potential in the video recognition domain. To optimize the speed-accuracy trade-off, we propose Semantic-aware Temporal Accumulation score (STA) to prune spatio-temporal tokens integrally. STA score considers two critical factors: temporal redundancy and semantic importance. The former depicts a specific region based on whether it is a new occurrence or a seen entity by aggregating token-to-token similarity in consecutive frames while the latter evaluates each token based on its contribution to the overall prediction. As a result, tokens with higher scores of STA carry more temporal redundancy as well as lower semantics thus being pruned. Based on the STA score, we are able to progressively prune the tokens without introducing any additional parameters or requiring further re-training. We directly apply the STA module to off-the-shelf ViT and VideoSwin backbones, and the empirical results on Kinetics-400 and Something-Something V2 achieve over 30% computation reduction with a negligible 0.2% accuracy drop. The code is released at https://github.com/Mark12Ding/STA.

Zijing Zhao, Sitong Wei, Qingchao Chen, Dehui Li, Yifan Yang, Yuxin Peng, Yang Liu

Domain adaptive Object Detection (DAOD) leverages a labeled domain (source) to learn an object detector generalizing to a novel domain without annotation (target). Recent advances use a teacher-student framework, i.e., a student model is supervised by the pseudo labels from a teacher model. Though great success, they suffer from the limited number of pseudo boxes with incorrect predictions caused by the domain shift, misleading the student model to get sub-optimal results. To mitigate this problem, we propose Masked Retraining Teacher-student framework (MRT) which leverages masked autoencoder and selective retraining mechanism on detection transformer. Specifically, we present a customized design of masked autoencoder branch, masking the multi-scale feature maps of target images and reconstructing features by the encoder of the student model and an auxiliary decoder. This helps the student model capture target domain characteristics and become a more data-efficient learner to gain knowledge from the limited number of pseudo boxes. Furthermore, we adopt selective retraining mechanism, periodically re-initializing certain parts of the student parameters with masked autoencoder refined weights to allow the model to jump out of the local optimum biased to the incorrect pseudo labels. Experimental results on three DAOD benchmarks demonstrate the effectiveness of our method. Code can be found at https://github.com/JeremyZhao1998/MRT-release.

Jiuming Liu, Guangming Wang, Zhe Liu, Chaokang Jiang, Marc Pollefeys, Hesheng Wang

Although point cloud registration has achieved remarkable advances in object-level and indoor scenes, large-scale registration methods are rarely explored. Challenges mainly arise from the huge point number, complex distribution, and outliers of outdoor LiDAR scans. In addition, most existing registration works generally adopt a two-stage paradigm: They first find correspondences by extracting discriminative local features and then leverage estimators (eg. RANSAC) to filter outliers, which are highly dependent on well-designed descriptors and post-processing choices. To address these problems, we propose an end-to-end transformer network (RegFormer) for large-scale point cloud alignment without any further post-processing. Specifically, a projection-aware hierarchical transformer is proposed to capture long-range dependencies and filter outliers by extracting point features globally. Our transformer has linear complexity, which guarantees high efficiency even for large-scale scenes. Furthermore, to effectively reduce mismatches, a bijective association transformer is designed for regressing the initial transformation. Extensive experiments on KITTI and NuScenes datasets demonstrate that our RegFormer achieves competitive performance in terms of both accuracy and efficiency. Codes are available at https://github.com/IRMVLab/RegFormer.

Ziming Chen, Yifeng Shi, Jinrang Jia

Cooperation between vehicles and infrastructure is vital to enhancing the safety of autonomous driving. Two significant and contradictory challenges now stand in the collaborative perception: fusion accuracy and communication bandwidth. Previous intermediate fusion methods that transmit features balance the accuracy and bandwidth compared with early fusion and late fusion, but usually have problems with feature alignment and domain gaps, and the bandwidth usage still falls short of the industrial application standard to our best knowledge. In this paper, we propose TransIFF, an instance-level feature fusion framework with transformers that can effectively reduce bandwidth usage. Furthermore, it can align the domain gaps between vehicle and infrastructure features, and improve the robustness of feature fusion, leading to a high cooperative perception accuracy. TransIFF is composed of three components: a vehicle-side network, an infrastructure-side network, and a vehicle-infrastructure fusion network. Initially, the vehicle-side and infrastructure-side networks independently generate instance-level features. Subsequently, the infrastructure-side instance-level features are transmitted to the vehicles, significantly reducing the communication bandwidth usage. Finally, in the vehicle-infrastructure fusion network, Cross-Domain Adaptation (CDA) module is designed to align the feature domains, followed by Feature Magnet (FM) module which can adaptively fuse the instance features and achieve a robust feature fusion. TransIFF yields state-of-the-art performance on the widely used real-world vehicle-infrastructure cooperative benchmark DAIR-V2X, achieving 59.62% AP with only 2^12 bytes bandwidth consumption.

Li Niu, Linfeng Tan, Xinhao Tao, Junyan Cao, Fengjun Guo, Teng Long, Liqing Zhang

Given a composite image, image harmonization aims to adjust the foreground illumination to be consistent with background. Previous methods have explored transforming foreground features to achieve competitive performance. In this work, we show that using global information to guide foreground feature transformation could achieve significant improvement. Besides, we propose to transfer the foreground-background relation from real images to composite images, which can provide intermediate supervision for the transformed encoder features. Additionally, considering the drawbacks of existing harmonization datasets, we also contribute a ccHarmony dataset which simulates the natural illumination variation. Extensive experiments on iHarmony4 and our contributed dataset demonstrate the superiority of our method. Our ccHarmony dataset is released at https://github.com/bcmi/Image-Harmonization-Dataset-ccHarmony.

Nicolas Aziere, Sinisa Todorovic

This paper addresses data augmentation for action segmentation. Our key novelty is that we augment the original training videos in the deep feature space, not in the visual spatiotemporal domain as done by previous work. For augmentation, we modify original deep features of video frames such that the resulting embeddings fall closer to the class decision boundaries. Also, we edit action sequences of the original training videos (a.k.a. transcripts) by inserting, deleting, and replacing actions such that the resulting transcripts are close in edit distance to the ground truth ones. For our data augmentation we resort to reinforcement learning, instead of more common supervised learning, since we do not have access to reliable oracles which would provide supervision about the optimal data modifications in the deep feature space. For modifying frame embeddings, we use a meta-model formulated as a Markov Game with multiple self-interested agents. Also, new transcripts are generated using a fast, parameter-free Monte Carlo tree search. Our experiments show that the proposed data augmentation of the Breakfast, GTEA, and 50Salads datasets leads to significant performance gains of several state of the art action segmenters.

Fangyun Wei, Yutong Chen

This work dedicates to continuous sign language recognition (CSLR), which is a weakly supervised task dealing with the recognition of continuous signs from videos, without any prior knowledge about the temporal boundaries between consecutive signs. Data scarcity heavily impedes the progress of CSLR. Existing approaches typically train CSLR models on a monolingual corpus, which is orders of magnitude smaller than that of speech recognition. In this work, we explore the feasibility of utilizing multilingual sign language corpora to facilitate monolingual CSLR. Our work is built upon the observation of cross-lingual signs, which originate from different sign languages but have similar visual signals (e.g., hand shape and motion). The underlying idea of our approach is to identify the cross-lingual signs in one sign language and properly leverage them as auxiliary training data to improve the recognition capability of another. To achieve the goal, we first build two sign language dictionaries containing isolated signs that appear in two datasets. Then we identify the sign-to-sign mappings between two sign languages via a well-optimized isolated sign language recognition model. At last, we train a CSLR model on the combination of the target data with original labels and the auxiliary data with mapped labels. Experimentally, our approach achieves state-of-the-art performance on two widely-used CSLR datasets: Phoenix-2014 and Phoenix-2014T.

Pin Tang, Hai-Ming Xu, Chao Ma

Knowledge transfer from multi-modal, i.e., LiDAR points and images, to a single LiDAR modal can take advantage of complimentary information from modal-fusion but keep a single modal inference speed, showing a promising direction for point cloud semantic segmentation in autonomous driving. Recent advances in point cloud segmentation distill knowledge from strictly aligned point-pixel fusion features while leaving a large number of unmatched image pixels unexplored and unmatched LiDAR points under-benefited. In this paper, we propose a novel approach, named ProtoTransfer, which not only fully exploits image representations but also transfers the learned multi-modal knowledge to all point cloud features. Specifically, based on the basic multi-modal learning framework, we build up a class-wise prototype bank from the strictly-aligned fusion features and encourage all the point cloud features to learn from the prototypes during model training. Moreover, to exploit the massive unmatched point and pixel features, we use a pseudo-labeling scheme and further accumulate these features into the class-wise prototype bank with a carefully designed fusion strategy. Without bells and whistles, our approach demonstrates superior performance over the published state-of-the-arts on two large-scale benchmarks, i.e., nuScenes and SemanticKITTI, and ranks 2nd on the competitive nuScenes Lidarseg challenge leaderboard.

Animesh Karnewar, Niloy J. Mitra, Andrea Vedaldi, David Novotny

Diffusion-based image generators can now produce high-quality and diverse samples, but their success has yet to fully translate to 3D generation: existing diffusion methods can either generate low-resolution but 3D consistent outputs, or detailed 2D views of 3D objects with potential structural defects and lacking either view consistency or realism. We present HoloFusion, a method that combines the best of these approaches to produce high-fidelity, plausible, and diverse 3D samples while learning from a collection of multi-view 2D images only. The method first generates coarse 3D samples using a variant of the recently proposed HoloDiffusion generator. Then, it independently renders and upsamples a large number of views of the coarse 3D model, super-resolves them to add detail, and distills those into a single, high-fidelity implicit 3D representation, which also ensures view-consistency of the final renders. The super-resolution network is trained as an integral part of HoloFusion, and the final distillation uses a new sampling scheme to capture the space of super-resolved signals. We compare our method against existing baselines, including DreamFusion, Get3D, EG3D, and HoloDiffusion, and achieve, to the best of our knowledge, the most realistic results on the challenging CO3Dv2 dataset.

Le Zhuo, Zhaokai Wang, Baisen Wang, Yue Liao, Chenxi Bao, Stanley Peng, Songhao Han, Aixi Zhang, Fei Fang, Si Liu

Music is essential when editing videos, but selecting music manually is difficult and time-consuming. Thus, we seek to automatically generate background music tracks given video input. This is a challenging task since it requires music-video datasets, efficient architectures for video-to-music generation, and reasonable metrics, none of which currently exist. To close this gap, we introduce a complete recipe including dataset, benchmark model, and evaluation metric for video background music generation. We present SymMV, a video and symbolic music dataset with various musical annotations. To the best of our knowledge, it is the first video-music dataset with rich musical annotations. We also propose a benchmark video background music generation framework named V-MusProd, which utilizes music priors of chords, melody, and accompaniment along with video-music relations of semantic, color, and motion features. To address the lack of objective metrics for video-music correspondence, we design a retrieval-based metric VMCP built upon a powerful video-music representation learning model. Experiments show that with our dataset, V-MusProd outperforms the state-of-the-art method in both music quality and correspondence with videos. We believe our dataset, benchmark model, and evaluation metric will boost the development of video background music generation. Our dataset and code are available at https://github.com/zhuole1025/SymMV.

Andong Deng, Taojiannan Yang, Chen Chen

The goal of building a benchmark (suite of datasets) is to provide a unified protocol for fair evaluation and thus facilitate the evolution of a specific area. Nonetheless, we point out that existing protocols of action recognition could yield partial evaluations due to several limitations. To comprehensively probe the effectiveness of spatiotemporal representation learning, we introduce BEAR, a new BEnchmark on video Action Recognition. BEAR is a collection of 18 video datasets grouped into 5 categories (anomaly, gesture, daily, sports, and instructional), which covers a diverse set of real-world applications. With BEAR, we thoroughly evaluate 6 common spatiotemporal models pre-trained by both supervised and self-supervised learning. We also report transfer performance via standard finetuning, few-shot fine-tuning, and unsupervised domain adaptation. Our observation suggests that the current state-of-the-art cannot solidly guarantee high performance on datasets close to real-world applications, and we hope BEAR can serve as a fair and challenging evaluation benchmark to gain insights on building next-generation spatiotemporal learners. Our dataset, code, and models are released at: https://github.com/AndongDeng/BEAR

Zhihong Chen, Shizhe Diao, Benyou Wang, Guanbin Li, Xiang Wan

Medical vision-and-language pre-training (Med-VLP) has shown promising improvements on many downstream medical tasks owing to its applicability to extracting generic representations from medical images and texts. Practically, there exist two typical types, i.e., the fusion-encoder type and the dual-encoder type, depending on whether a heavy fusion module is used. The former is superior at multi-modal tasks owing to the sufficient interaction between modalities; the latter is good at uni-modal and cross-modal tasks due to the single-modality encoding ability. To take advantage of these two types, we propose an effective yet straightforward scheme named PTUnifier to unify the two types. We first unify the input format by introducing visual and textual prompts, which serve as DETR-like queries that assist in extracting features when one of the modalities is missing. By doing so, a single model could serve as a foundation model that processes various tasks adopting different input formats (i.e., image-only, text-only, and image-text-pair). Furthermore, we construct a prompt pool (instead of static ones) to improve diversity and scalability, enabling queries conditioned on different input instances. Experimental results show that our approach achieves state-of-the-art results on a broad range of tasks, spanning uni-modal tasks (i.e., image/text classification and text summarization), cross-modal tasks (i.e., image-to-text generation and image-text/text-image retrieval), and multi-modal tasks (i.e., visual question answering), demonstrating the effectiveness of our approach. Note that the adoption of prompts is orthogonal to most existing Med-VLP approaches and could be a beneficial and complementary extension to these approaches. The source code is available at https://anonymous.4open.science/r/ICCV-2023-Submission-PTUnifier/ and will be released in the final version of this paper.

Sucheng Ren, Xingyi Yang, Songhua Liu, Xinchao Wang

Vision Transformer has demonstrated impressive success across various vision tasks. However, its heavy computation cost, which grows quadratically with respect to the token sequence length, largely limits its power in handling large feature maps. To alleviate the computation cost, previous works rely on either fine-grained self-attentions restricted to local small regions, or global self-attentions but to shorten the sequence length resulting in coarse granularity. In this paper, we propose a novel model, termed as Self-guided Transformer (SG-Former), towards effective global self-attention with adaptive fine granularity. At the heart of our approach is to utilize a significance map, which is estimated through hybrid-scale self-attention and evolves itself during training, to reallocate tokens based on the significance of each region. Intuitively, we assign more tokens to the salient regions for achieving fine-grained attention, while allocating fewer tokens to the minor regions in exchange for efficiency and global receptive fields. The proposed SG-Former achieves performance superior to state of the art: our base size model achieves 84.7% Top-1 accuracy on ImageNet-1K, 51.2mAP bbAP on CoCo, 52.7mIoU on ADE20K surpassing the Swin Transformer by +1.3% / +2.7 mAP/ +3 mIoU, with lower computation costs and fewer parameters.