CVPR 2022 论文列表

This paper aims at recovering the shape of a scene with unknown, non-Lambertian, and possibly spatially-varying surface materials. When the shape of the object is highly complex and that shadows cast on the surface, the task becomes very challenging. To overcome these challenges, we propose a coordinate-based deep MLP (multilayer perceptron) to parameterize both the unknown 3D shape and the unknown reflectance at every surface point. This network is able to leverage the observed photometric variance and shadows on the surface, and recover both surface shape and general non-Lambertian reflectance. We explicitly predict cast shadows, mitigating possible artifacts on these shadowing regions, leading to higher estimation accuracy. Our framework is entirely self-supervised, in the sense that it requires neither ground truth shape nor BRDF. Tests on real-world images demonstrate that our method outperform existing methods by a significant margin. Thanks to the small size of the MLP-net, our method is an order of magnitude faster than previous CNN-based methods.

In this paper, we introduce a novel deep learning method for photo-realistic manipulation of the emotional state of actors in "in-the-wild" videos. The proposed method is based on a parametric 3D face representation of the actor in the input scene that offers a reliable disentanglement of the facial identity from the head pose and facial expressions. It then uses a novel deep domain translation framework that alters the facial expressions in a consistent and plausible manner, taking into account their dynamics. Finally, the altered facial expressions are used to photo-realistically manipulate the facial region in the input scene based on an especially-designed neural face renderer. To the best of our knowledge, our method is the first to be capable of controlling the actor's facial expressions by even using as a sole input the semantic labels of the manipulated emotions, while at the same time preserving the speech-related lip movements. We conduct extensive qualitative and quantitative evaluations and comparisons, which demonstrate the effectiveness of our approach and the especially promising results that we obtain. Our method opens a plethora of new possibilities for useful applications of neural rendering technologies, ranging from movie post-production and video games to photo-realistic affective avatars.

We propose a novel approach for 3D video synthesis that is able to represent multi-view video recordings of a dynamic real-world scene in a compact, yet expressive representation that enables high-quality view synthesis and motion interpolation. Our approach takes the high quality and compactness of static neural radiance fields in a new direction: to a model-free, dynamic setting. At the core of our approach is a novel time-conditioned neural radiance field that represents scene dynamics using a set of compact latent codes. We are able to significantly boost the training speed and perceptual quality of the generated imagery by a novel hierarchical training scheme in combination with ray importance sampling. Our learned representation is highly compact and able to represent a 10 second 30 FPS multi-view video recording by 18 cameras with a model size of only 28MB. We demonstrate that our method can render high-fidelity wide-angle novel views at over 1K resolution, even for complex and dynamic scenes. We perform an extensive qualitative and quantitative evaluation that shows that our approach outperforms the state of the art. Project website: https://neural-3d-video.github.io/.

This paper addresses the challenge of reconstructing 3D indoor scenes from multi-view images. Many previous works have shown impressive reconstruction results on textured objects, but they still have difficulty in handling low-textured planar regions, which are common in indoor scenes. An approach to solving this issue is to incorporate planer constraints into the depth map estimation in multi-view stereo-based methods, but the per-view plane estimation and depth optimization lack both efficiency and multi-view consistency. In this work, we show that the planar constraints can be conveniently integrated into the recent implicit neural representation-based reconstruction methods. Specifically, we use an MLP network to represent the signed distance function as the scene geometry. Based on the Manhattan-world assumption, planar constraints are employed to regularize the geometry in floor and wall regions predicted by a 2D semantic segmentation network. To resolve the inaccurate segmentation, we encode the semantics of 3D points with another MLP and design a novel loss that jointly optimizes the scene geometry and semantics in 3D space. Experiments on ScanNet and 7-Scenes datasets show that the proposed method outperforms previous methods by a large margin on 3D reconstruction quality. The code and supplementary materials are available at https://zju3dv.github.io/manhattan_sdf.

While NeRF has shown great success for neural reconstruction and rendering, its limited MLP capacity and long per-scene optimization times make it challenging to model large-scale indoor scenes. In contrast, classical 3D reconstruction methods can handle large-scale scenes but do not produce realistic renderings. We propose NeRFusion, a method that combines the advantages of NeRF and TSDF-based fusion techniques to achieve efficient large-scale reconstruction and photo-realistic rendering. We process the input image sequence to predict per-frame local radiance fields via direct network inference. These are then fused using a novel recurrent neural network that incrementally reconstructs a global, sparse scene representation in real-time. This global volume can be further fine-tuned to boost rendering quality. We demonstrate that NeRFusion achieves state-of-the-art quality on both large-scale indoor and small-scale object scenes, with substantially faster reconstruction than NeRF and other recent methods.

Neural Radiance Fields (NeRF) is a technique for high quality novel view synthesis from a collection of posed input images. Like most view synthesis methods, NeRF uses tonemapped low dynamic range (LDR) as input; these images have been processed by a lossy camera pipeline that smooths detail, clips highlights, and distorts the simple noise distribution of raw sensor data. We modify NeRF to instead train directly on linear raw images, preserving the scene's full dynamic range. By rendering raw output images from the resulting NeRF, we can perform novel high dynamic range (HDR) view synthesis tasks. In addition to changing the camera viewpoint, we can manipulate focus, exposure, and tonemapping after the fact. Although a single raw image appears significantly more noisy than a postprocessed one, we show that NeRF is highly robust to the zero-mean distribution of raw noise. When optimized over many noisy raw inputs (25-200), NeRF produces a scene representation so accurate that its rendered novel views outperform dedicated single and multi-image deep raw denoisers run on the same wide baseline input images. As a result, our method, which we call RawNeRF, can reconstruct scenes from extremely noisy images captured in near-darkness.

Natural language explanation (NLE) models aim at explaining the decision-making process of a black box system via generating natural language sentences which are human-friendly, high-level and fine-grained. Current NLE models explain the decision-making process of a vision or vision-language model (a.k.a., task model), e.g., a VQA model, via a language model (a.k.a., explanation model), e.g., GPT. Other than the additional memory resources and inference time required by the task model, the task and explanation models are completely independent, which disassociates the explanation from the reasoning process made to predict the answer. We introduce NLX-GPT, a general, compact and faithful language model that can simultaneously predict an answer and explain it. We first conduct pre-training on large scale data of image-caption pairs for general understanding of images, and then formulate the answer as a text prediction task along with the explanation. Without region proposals nor a task model, our resulting overall framework attains better evaluation scores, contains much less parameters and is 15x faster than the current SoA model. We then address the problem of evaluating the explanations which can be in many times generic, data-biased and can come in several forms. We therefore design 2 new evaluation measures: (1) explain-predict and (2) retrieval-based attack, a self-evaluation framework that requires no labels. Code is at: https://github.com/cvpr2022annonymous/nlxgpt

Multi-view depth estimation methods typically require the computation of a multi-view cost-volume, which leads to huge memory consumption and slow inference. Furthermore, multi-view matching can fail for texture-less surfaces, reflective surfaces and moving objects. For such failure modes, single-view depth estimation methods are often more reliable. To this end, we propose MaGNet, a novel framework for fusing single-view depth probability with multi-view geometry, to improve the accuracy, robustness and efficiency of multi-view depth estimation. For each frame, MaGNet estimates a single-view depth probability distribution, parameterized as a pixel-wise Gaussian. The distribution estimated for the reference frame is then used to sample per-pixel depth candidates. Such probabilistic sampling enables the network to achieve higher accuracy while evaluating fewer depth candidates. We also propose depth consistency weighting for the multi-view matching score, to ensure that the multi-view depth is consistent with the single-view predictions. The proposed method achieves state-of-the-art performance on ScanNet, 7-Scenes and KITTI. Qualitative evaluation demonstrates that our method is more robust against challenging artifacts such as texture-less/reflective surfaces and moving objects. Our code and model weights are available at https://github.com/baegwangbin/MaGNet.

Obtaining accurate and diverse human motion prediction is essential to many industrial applications, especially robotics and autonomous driving. Recent research has explored several techniques to enhance diversity and maintain the accuracy of human motion prediction at the same time. However, most of them need to define a combined loss, such as the weighted sum of accuracy loss and diversity loss, and then decide their weights as hyperparameters before training. In this work, we aim to design a prediction framework that can balance the accuracy sampling and diversity sampling during the testing phase. In order to achieve this target, we propose a multi-objective conditional variational inference prediction model. We also propose a short-term oracle to encourage the prediction framework to explore more diverse future motions. We evaluate the performance of our proposed approach on two standard human motion datasets. The experiment results show that our approach is effective and on a par with state-of-the-art performance in terms of accuracy and diversity.

Multi-view clustering can explore common semantics from multiple views and has attracted increasing attention. However, existing works punish multiple objectives in the same feature space, where they ignore the conflict between learning consistent common semantics and reconstructing inconsistent view-private information. In this paper, we propose a new framework of multi-level feature learning for contrastive multi-view clustering to address the aforementioned issue. Our method learns different levels of features from the raw features, including low-level features, high-level features, and semantic labels/features in a fusion-free manner, so that it can effectively achieve the reconstruction objective and the consistency objectives in different feature spaces. Specifically, the reconstruction objective is conducted on the low-level features. Two consistency objectives based on contrastive learning are conducted on the high-level features and the semantic labels, respectively. They make the high-level features effectively explore the common semantics and the semantic labels achieve the multi-view clustering. As a result, the proposed framework can reduce the adverse influence of view-private information. Extensive experiments on public datasets demonstrate that our method achieves state-of-the-art clustering effectiveness.

A novel ordinal regression algorithm, called moving window regression (MWR), is proposed in this paper. First, we propose the notion of relative rank (rho-rank), which is a new order representation scheme for input and reference instances. Second, we develop global and local relative regressors (rho-regressors) to predict rho-ranks within entire and specific rank ranges, respectively. Third, we refine an initial rank estimate iteratively by selecting two reference instances to form a search window and then estimating the rho-rank within the window. Extensive experiments results show that the proposed algorithm achieves the state-of-the-art performances on various benchmark datasets for facial age estimation and historical color image classification. The codes are available at https://github.com/nhshin-mcl/MWR.

Implicit neural representation (INR) has been successful in representing static images. Contemporary image-based INR, with the use of Fourier-based positional encoding, can be viewed as a mapping from sinusoidal patterns with different frequencies to image content. Inspired by that view, we hypothesize that it is possible to generate temporally varying content with a single image-based INR model by displacing its input sinusoidal patterns over time. By exploiting the relation between the phase information in sinusoidal functions and their displacements, we incorporate into the conventional image-based INR model a phase-varying positional encoding module, and couple it with a phase-shift generation module that determines the phase-shift values at each frame. The model is trained end-to-end on a video to jointly determine the phase-shift values at each time with the mapping from the phase-shifted sinusoidal functions to the corresponding frame, enabling an implicit video representation. Experiments on a wide range of videos suggest that such a model is capable of learning to interpret phase-varying positional embeddings into the corresponding time-varying content. More importantly, we found that the learned phase-shift vectors tend to capture meaningful temporal and motion information from the video. In particular, manipulating the phase-shift vectors induces meaningful changes in the temporal dynamics of the resulting video, enabling non-trivial temporal and motion editing effects such as temporal interpolation, motion magnification, motion smoothing, and video loop detection.

We present Mobile-Former, a parallel design of MobileNet and transformer with a two-way bridge in between. This structure leverages the advantages of MobileNet at local processing and transformer at global interaction. And the bridge enables bidirectional fusion of local and global features. Different from recent works on vision transformer, the transformer in Mobile-Former contains very few tokens (e.g. 6 or fewer tokens) that are randomly initialized to learn global priors, resulting in low computational cost. Combining with the proposed light-weight cross attention to model the bridge, Mobile-Former is not only computationally efficient, but also has more representation power. It outperforms MobileNetV3 at low FLOP regime from 25M to 500M FLOPs on ImageNet classification. For instance, Mobile-Former achieves 77.9% top-1 accuracy at 294M FLOPs, gaining 1.3% over MobileNetV3 but saving 17% of computations. When transferring to object detection, Mobile-Former outperforms MobileNetV3 by 8.6 AP in RetinaNet framework. Furthermore, we build an efficient end-to-end detector by replacing backbone, encoder and decoder in DETR with Mobile-Former, which outperforms DETR by 1.3 AP but saves 52% of computational cost and 36% of parameters. Code will be released at https://github.com/aaboys/mobileformer.

We introduce AdaMix, an adaptive differentially private algorithm for training deep neural network classifiers using both private and public image data. While pre-training language models on large public datasets has enabled strong differential privacy (DP) guarantees with minor loss of accuracy, a similar practice yields punishing trade-offs in vision tasks. A few-shot or even zero-shot learning baseline that ignores private data can outperform fine-tuning on a large private dataset. AdaMix incorporates few-shot training, or cross-modal zero-shot learning, on public data prior to private fine-tuning, to improve the trade-off. AdaMix reduces the error increase from the non-private upper bound from the 167-311% of the baseline, on average across 6 datasets, to 68-92% depending on the desired privacy level selected by the user. AdaMix tackles the trade-off arising in visual classification, whereby the most privacy sensitive data, corresponding to isolated points in representation space, are also critical for high classification accuracy. In addition, AdaMix comes with strong theoretical privacy guarantees and convergence analysis.

While local-window self-attention performs notably in vision tasks, it suffers from limited receptive field and weak modeling capability issues. This is mainly because it performs self-attention within non-overlapped windows and shares weights on the channel dimension. We propose MixFormer to find a solution. First, we combine local-window self-attention with depth-wise convolution in a parallel design, modeling cross-window connections to enlarge the receptive fields. Second, we propose bi-directional interactions across branches to provide complementary clues in the channel and spatial dimensions. These two designs are integrated to achieve efficient feature mixing among windows and dimensions. Our MixFormer provides competitive results on image classification with EfficientNet and shows better results than RegNet and Swin Transformer. Performance in downstream tasks outperforms its alternatives by significant margins with less computational costs in 5 dense prediction tasks on MS COCO, ADE20k, and LVIS. Code is available at https://github.com/PaddlePaddle/PaddleClas.

Tracking often uses a multi-stage pipeline of feature extraction, target information integration, and bounding box estimation. To simplify this pipeline and unify the process of feature extraction and target information integration, we present a compact tracking framework, termed as MixFormer, built upon transformers. Our core design is to utilize the flexibility of attention operations, and propose a Mixed Attention Module (MAM) for simultaneous feature extraction and target information integration. This synchronous modeling scheme allows to extract target-specific discriminative features and perform extensive communication between target and search area. Based on MAM, we build our MixFormer tracking framework simply by stacking multiple MAMs with progressive patch embedding and placing a localization head on top. In addition, to handle multiple target templates during online tracking, we devise an asymmetric attention scheme in MAM to reduce computational cost, and propose an effective score prediction module to select high-quality templates. Our MixFormer sets a new state-of-the-art performance on five tracking benchmarks, including LaSOT, TrackingNet, VOT2020, GOT-10k, and UAV123. In particular, our MixFormer-L achieves NP score of 79.9% on LaSOT, 88.9% on TrackingNet and EAO of 0.555 on VOT2020. We also perform in-depth ablation studies to demonstrate the effectiveness of simultaneous feature extraction and information integration. Code and trained models are publicly available at https://github.com/MCG-NJU/MixFormer.

Though neural radiance fields ("NeRF") have demonstrated impressive view synthesis results on objects and small bounded regions of space, they struggle on "unbounded" scenes, where the camera may point in any direction and content may exist at any distance. In this setting, existing NeRF-like models often produce blurry or low-resolution renderings (due to the unbalanced detail and scale of nearby and distant objects), are slow to train, and may exhibit artifacts due to the inherent ambiguity of the task of reconstructing a large scene from a small set of images. We present an extension of mip-NeRF (a NeRF variant that addresses sampling and aliasing) that uses a non-linear scene parameterization, online distillation, and a novel distortion-based regularizer to overcome the challenges presented by unbounded scenes. Our model, which we dub "mip-NeRF 360" as we target scenes in which the camera rotates 360 degrees around a point, reduces mean-squared error by 57% compared to mip-NeRF, and is able to produce realistic synthesized views and detailed depth maps for highly intricate, unbounded real-world scenes.

Transformers have shown great potential in computer vision tasks. A common belief is their attention-based token mixer module contributes most to their competence. However, recent works show the attention-based module in transformers can be replaced by spatial MLPs and the resulted models still perform quite well. Based on this observation, we hypothesize that the general architecture of the transformers, instead of the specific token mixer module, is more essential to the model's performance. To verify this, we deliberately replace the attention module in transformers with an embarrassingly simple spatial pooling operator to conduct only basic token mixing. Surprisingly, we observe that the derived model, termed as PoolFormer, achieves competitive performance on multiple computer vision tasks. For example, on ImageNet-1K, PoolFormer achieves 82.1% top-1 accuracy, surpassing well-tuned vision transformer/MLP-like baselines DeiT-B/ResMLP-B24 by 0.3%/1.1% accuracy with 35%/52% fewer parameters and 49%/61% fewer MACs. The effectiveness of PoolFormer verifies our hypothesis and urges us to initiate the concept of "MetaFormer", a general architecture abstracted from transformers without specifying the token mixer. Based on the extensive experiments, we argue that MetaFormer is the key player in achieving superior results for recent transformer and MLP-like models on vision tasks. This work calls for more future research dedicated to improving MetaFormer instead of focusing on the token mixer modules. Additionally, our proposed PoolFormer could serve as a starting baseline for future MetaFormer architecture design.

The existing pose estimation approaches often require a large number of annotated images to attain good estimation performance, which are laborious to acquire. To reduce the human efforts on pose annotations, we propose a novel Meta Agent Teaming Active Learning (MATAL) framework to actively select and label informative images for effective learning. Our MATAL formulates the image selection procedure as a Markov Decision Process and learns an optimal sampling policy that directly maximizes the performance of the pose estimator. Our framework consists of a novel state-action representation as well as a multi-agent team to enable batch sampling in the active learning procedure. The framework could be effectively optimized via Meta-Optimization to accelerate the adaptation to the gradually expanded labeled data during deployment. Finally, we show experimental results on both human hand and body pose estimation benchmark datasets and demonstrate that our method significantly outperforms all baselines continuously under the same amount of annotation budget. Moreover, to obtain similar pose estimation accuracy, our MATAL framework can save around 40% labeling efforts on average compared to state-of-the-art active learning frameworks.