CVPR 2022 论文列表

The size and shape of the receptive field determine how the network aggregates local features, and affect the overall performance of a model considerably. Many components in a neural network, such as depth, kernel sizes, and strides for convolution and pooling, influence the receptive field. However, they still rely on hyperparameters, and the receptive fields of existing models result in suboptimal shapes and sizes. Hence, we propose a simple yet effective Dynamically Optimized Pooling operation, referred to as DynOPool, which learns the optimized scale factors of feature maps end-to-end. Moreover, DynOPool determines the proper resolution of a feature map by learning the desirable size and shape of its receptive field, which allows an operator in a deeper layer to observe an input image in the optimal scale. Any kind of resizing modules in a deep neural network can be replaced by DynOPool with minimal cost. Also, DynOPool controls the complexity of the model by introducing an additional loss term that constrains computational cost. Our experiments show that the models equipped with the proposed learnable resizing module outperform the baseline algorithms on multiple datasets in image classification and semantic segmentation.

While most state-of-the-art instance segmentation methods produce binary segmentation masks, geographic and cartographic applications typically require precise vector polygons of extracted objects instead of rasterized output. This paper introduces PolyWorld, a neural network that directly extracts building vertices from an image and connects them correctly to create precise polygons. The model predicts the connection strength between each pair of vertices using a graph neural network and estimates the assignments by solving a differentiable optimal transport problem. Moreover, the vertex positions are optimized by minimizing a combined segmentation and polygonal angle difference loss. PolyWorld significantly outperforms the state of the art in building polygonization and achieves not only notable quantitative results, but also produces visually pleasing building polygons. Code and trained weights are publicly available at https://github.com/zorzi-s/PolyWorldPretrainedNetwork.

Optimization of Top-1 ImageNet promotes enormous networks that may be impractical in inference settings. Binary neural networks (BNNs) have the potential to significantly lower the compute intensity but existing models suffer from low quality. To overcome this deficiency, we propose PokeConv, a binary convolution block which improves quality of BNNs by techniques such as adding multiple residual paths, and tuning the activation function. We apply it to ResNet-50 and optimize ResNet's initial convolutional layer which is hard to binarize. We name the resulting network family PokeBNN. These techniques are chosen to yield favorable improvements in both top-1 accuracy and the network's cost. In order to enable joint optimization of the cost together with accuracy, we define arithmetic computation effort (ACE), a hardware- and energy-inspired cost metric for quantized and binarized networks. We also identify a need to optimize an under-explored hyper-parameter controlling the binarization gradient approximation. We establish a new, strong state-of-the-art (SOTA) on top-1 accuracy together with commonly-used CPU64 cost, ACE cost and network size metrics. ReActNet-Adam, the previous SOTA in BNNs, achieved a 70.5% top-1 accuracy with 7.9 ACE. A small variant of PokeBNN achieves 70.5% top-1 with 2.6 ACE, more than 3x reduction in cost; a larger PokeBNN achieves 75.6% top-1 with 7.8 ACE, more than 5% improvement in accuracy without increasing the cost. PokeBNN implementation in JAX / Flax and reproduction instructions are open sourced. Source code and reproduction instructions are available in AQT repository: https://github.com/google/aqt.

Recently, zero-shot and few-shot learning via Contrastive Vision-Language Pre-training (CLIP) have shown inspirational performance on 2D visual recognition, which learns to match images with their corresponding texts in open-vocabulary settings. However, it remains under explored that whether CLIP, pre-trained by large-scale image-text pairs in 2D, can be generalized to 3D recognition. In this paper, we identify such a setting is feasible by proposing PointCLIP, which conducts alignment between CLIP-encoded point cloud and 3D category texts. Specifically, we encode a point cloud by projecting it into multi-view depth maps without rendering, and aggregate the view-wise zero-shot prediction to achieve knowledge transfer from 2D to 3D. On top of that, we design an inter-view adapter to better extract the global feature and adaptively fuse the few-shot knowledge learned from 3D into CLIP pre-trained in 2D. By just fine-tuning the lightweight adapter in the few-shot settings, the performance of PointCLIP could be largely improved. In addition, we observe the complementary property between PointCLIP and classical 3D-supervised networks. By simple ensembling, PointCLIP boosts baseline's performance and even surpasses state-of-the-art models. Therefore, PointCLIP is a promising alternative for effective 3D point cloud understanding via CLIP under low resource cost and data regime. We conduct thorough experiments on widely-adopted ModelNet10, ModelNet40 and the challenging ScanObjectNN to demonstrate the effectiveness of PointCLIP.

We present a simple and effective framework, named Point2Seq, for 3D object detection from point clouds. In contrast to previous methods that normally predict attributes of 3D objects all at once, we expressively model the interdependencies between attributes of 3D objects, which in turn enables a better detection accuracy. Specifically, we view each 3D object as a sequence of words and reformulate the 3D object detection task as decoding words from 3D scenes in an auto-regressive manner. We further propose a lightweight scene-to-sequence decoder that can auto-regressively generate words conditioned on features from a 3D scene as well as cues from the preceding words. The predicted words eventually constitute a set of sequences that completely describe the 3D objects in the scene, and all the predicted sequences are then automatically assigned to the respective ground truths through similarity-based sequence matching. Our approach is conceptually intuitive and can be readily plugged upon most existing 3D-detection backbones without adding too much computational overhead; the sequential decoding paradigm we proposed, on the other hand, can better exploit information from complex 3D scenes with the aid of preceding predicted words. Without bells and whistles, our method significantly outperforms the previous anchor- and center-based 3D object detection frameworks, yielding the new state-of-the-art on the challenging ONCE dataset as well as the Waymo Open Dataset. Code will be made publicly available.

We propose Point2Cyl, a supervised network transforming a raw 3D point cloud to a set of extrusion cylinders. Reverse engineering from a raw geometry to a CAD model is an essential task to enable manipulation of the 3D data in shape editing software and thus expand their usages in many downstream applications. Particularly, the form of CAD models having a sequence of extrusion cylinders --- a 2D sketch plus an extrusion axis and range --- and their boolean combinations is not only widely used in the CAD community/software but also has great expressivity of shapes, compared to having limited types of primitives (e.g., planes, spheres, and cylinders). In this work, we introduce a neural network that solves the extrusion cylinder decomposition problem in a geometry-grounded way by first learning underlying geometric proxies. Precisely, our approach first predicts per-point segmentation, base/barrel labels and normals, then estimates for the underlying extrusion parameters in differentiable and closed-form formulations. Our experiments show that our approach demonstrates the best performance on two recent CAD datasets, Fusion Gallery and DeepCAD, and we further showcase our approach on reverse engineering and editing.

This article addresses the problem of distilling knowledge from a large teacher model to a slim student network for LiDAR semantic segmentation. Directly employing previous distillation approaches yields inferior results due to the intrinsic challenges of point cloud, i.e., sparsity, randomness and varying density. To tackle the aforementioned problems, we propose the Point-to-Voxel Knowledge Distillation (PVD), which transfers the hidden knowledge from both point level and voxel level. Specifically, we first leverage both the pointwise and voxelwise output distillation to complement the sparse supervision signals. Then, to better exploit the structural information, we divide the whole point cloud into several supervoxels and design a difficultyaware sampling strategy to more frequently sample supervoxels containing less frequent classes and faraway objects. On these supervoxels, we propose inter-point and intervoxel affinity distillation, where the similarity information between points and voxels can help the student model better capture the structural information of the surrounding environment. We conduct extensive experiments on two popular LiDAR segmentation benchmarks, i.e., nuScenes [3] and SemanticKITTI [1]. On both benchmarks, our PVD consistently outperforms previous distillation approaches by a large margin on three representative backbones, i.e., Cylinder3D [27, 28], SPVNAS [20] and MinkowskiNet [5]. Notably, on the challenging nuScenes and SemanticKITTI datasets, our method can achieve roughly 75% MACs reduction and 2x speedup on the competitive Cylinder3D model and rank 1st on the SemanticKITTI leaderboard among all published algorithms. Our code is available at https://github.com/cardwing/Codes-for-PVKD.

We present Point-BERT, a novel paradigm for learning Transformers to generalize the concept of BERT onto 3D point cloud. Following BERT, we devise a Masked Point Modeling (MPM) task to pre-train point cloud Transformers. Specifically, we first divide a point cloud into several local patches, and a point cloud Tokenizer is devised via a discrete Variational AutoEncoder (dVAE) to generate discrete point tokens containing meaningful local information. Then, we randomly mask some patches of input point clouds and feed them into the backbone Transformer. The pre-training objective is to recover the original point tokens at the masked locations under the supervision of point tokens obtained by the Tokenizer. Extensive experiments demonstrate that the proposed BERT-style pre-training strategy significantly improves the performance of standard point cloud Transformers. Equipped with our pre-training strategy, we show that a pure Transformer architecture attains 93.7% accuracy on ModelNet40 and 83.1% accuracy on the hardest setting of ScanObjectNN, surpassing carefully designed point cloud models with much fewer hand-made designs and human priors. We also demonstrate that the representations learned by Point-BERT transfer well to new tasks and domains, where our models largely advance the state-of-the-art of few-shot point cloud classification task.

LiDAR has become one of the primary 3D object detection sensors in autonomous driving. However, LiDAR's diverging point pattern with increasing distance results in a non-uniform sampled point cloud ill-suited to discretized volumetric feature extraction. Current methods either rely on voxelized point clouds or use inefficient farthest point sampling to mitigate detrimental effects caused by density variation but largely ignore point density as a feature and its predictable relationship with distance from the LiDAR sensor. Our proposed solution, Point Density-Aware Voxel network (PDV), is an end-to-end two stage LiDAR 3D object detection architecture that is designed to account for these point density variations. PDV efficiently localizes voxel features from the 3D sparse convolution backbone through voxel point centroids. The spatially localized voxel features are then aggregated through a density-aware RoI grid pooling module using kernel density estimation (KDE) and self-attention with point density positional encoding. Finally, we exploit LiDAR's point density to distance relationship to refine our final bounding box confidences. PDV outperforms all state-of-the-art methods on the Waymo Open Dataset and achieves competitive results on the KITTI dataset.

The construction of 3D point cloud datasets requires a great deal of human effort. Therefore, constructing a largescale 3D point clouds dataset is difficult. In order to remedy this issue, we propose a newly developed point cloud fractal database (PC-FractalDB), which is a novel family of formula-driven supervised learning inspired by fractal geometry encountered in natural 3D structures. Our research is based on the hypothesis that we could learn representations from more real-world 3D patterns than conventional 3D datasets by learning fractal geometry. We show how the PC-FractalDB facilitates solving several recent dataset-related problems in 3D scene understanding, such as 3D model collection and labor-intensive annotation. The experimental section shows how we achieved the performance rate of up to 61.9% and 59.0% for the ScanNetV2 and SUN RGB-D datasets, respectively, over the current highest scores obtained with the PointContrast, contrastive scene contexts (CSC), and RandomRooms. Moreover, the PC-FractalDB pre-trained model is especially effective in training with limited data. For example, in 10% of training data on ScanNetV2, the PC-FractalDB pre-trained VoteNet performs at 38.3%, which is +14.8% higher accuracy than CSC. Of particular note, we found that the proposed method achieves the highest results for 3D object detection pre-training in limited point cloud data.

In this paper, we present Point Cloud Color Constancy, in short PCCC, an illumination chromaticity estimation algorithm exploiting a point cloud. We leverage the depth information captured by the time-of-flight (ToF) sensor mounted rigidly with the RGB sensor, and form a 6D cloud where each point contains the coordinates and RGB intensities, noted as (x,y,z,r,g,b). PCCC applies the PointNet architecture to the color constancy problem, deriving the illumination vector point-wise and then making a global decision about the global illumination chromaticity. On two popular RGB-D datasets, which we extend with illumination information, as well as on a novel benchmark, PCCC obtains lower error than the state-of-the-art algorithms. Our method is simple and fast, requiring merely 16*16-size input and reaching speed over 500 fps, including the cost of building the point cloud and net inference.

We present Playable Environments - a new representation for interactive video generation and manipulation in space and time. With a single image at inference time, our novel framework allows the user to move objects in 3D while generating a video by providing a sequence of desired actions. The actions are learnt in an unsupervised manner. The camera can be controlled to get the desired viewpoint. Our method builds an environment state for each frame, which can be manipulated by our proposed action module and decoded back to the image space with volumetric rendering. To support diverse appearances of objects, we extend neural radiance fields with style-based modulation. Our method trains on a collection of various monocular videos requiring only the estimated camera parameters and 2D object locations. To set a challenging benchmark, we introduce two large scale video datasets with significant camera movements. As evidenced by our experiments, playable environments enable several creative applications not attainable by prior video synthesis works, including playable 3D video generation, stylization and manipulation. We make our code, demos and datasets publicly available.

We present a novel framework named PlaneMVS for 3D plane reconstruction from multiple input views with known camera poses. Most previous learning-based plane reconstruction methods reconstruct 3D planes from single images, which highly rely on single-view regression and suffer from depth scale ambiguity. In contrast, we reconstruct 3D planes with a multi-view-stereo (MVS) pipeline that takes advantage of multi-view geometry. We decouple plane reconstruction into a semantic plane detection branch and a plane MVS branch. The semantic plane detection branch is based on a single-view plane detection framework but with differences. The plane MVS branch adopts a set of slanted plane hypotheses to replace conventional depth hypotheses to perform plane sweeping strategy and finally learns pixel-level plane parameters and its planar depth map. We present how the two branches are learned in a balanced way, and propose a soft-pooling loss to associate the outputs of the two branches and make them benefit from each other. Extensive experiments on various indoor datasets show that PlaneMVS significantly outperforms state-of-the-art (SOTA) single-view plane reconstruction methods on both plane detection and 3D geometry metrics. Our method even outperforms a set of SOTA learning-based MVS methods thanks to the learned plane priors. To the best of our knowledge, this is the first work on 3D plane reconstruction within an end-to-end MVS framework.

We present PlanarRecon -- a novel framework for globally coherent detection and reconstruction of 3D planes from a posed monocular video. Unlike previous works that detect planes in 2D from a single image, PlanarRecon incrementally detects planes in 3D for each video fragment, which consists of a set of key frames, from a volumetric representation of the scene using neural networks. A learning-based tracking and fusion module is designed to merge planes from previous fragments to form a coherent global plane reconstruction. Such design allows PlanarRecon to integrate observations from multiple views within each fragment and temporal information across different ones, resulting in an accurate and coherent reconstruction of the scene abstraction with low-polygonal geometry. Experiments show that the proposed approach achieves state-of-the-art performances on the ScanNet dataset while being real-time.

Blind deblurring has attracted much interest with its wide applications in reality. The blind deblurring problem is usually solved by estimating the intermediate kernel and the intermediate image alternatively, which will finally converge to the blurring kernel of the observed image. Numerous works have been proposed to obtain intermediate images with fewer undesirable artifacts by designing delicate regularization on the latent solution. However, these methods still fail while dealing with images containing saturations and large blurs. To address this problem, we propose an intermediate image correction method which utilizes Bayes posterior estimation to screen through the intermediate image and exclude those unfavorable pixels to reduce their influence for kernel estimation. Extensive experiments have proved that the proposed method can effectively improve the accuracy of the final derived kernel against the state-of-the-art methods on benchmark datasets by both quantitative and qualitative comparisons.

In real-world applications of machine learning, reliable and safe systems must consider measures of performance beyond standard test set accuracy. These other goals include out-of-distribution (OOD) robustness, prediction consistency, resilience to adversaries, calibrated uncertainty estimates, and the ability to detect anomalous inputs. However, improving performance towards these goals is often a balancing act that today's methods cannot achieve without sacrificing performance on other safety axes. For instance, adversarial training improves adversarial robustness but sharply degrades other classifier performance metrics. Similarly, strong data augmentation and regularization techniques often improve OOD robustness but harm anomaly detection, raising the question of whether a Pareto improvement on all existing safety measures is possible. To meet this challenge, we design a new data augmentation strategy utilizing the natural structural complexity of pictures such as fractals, which outperforms numerous baselines, is near Pareto-optimal, and roundly improves safety measures.

We propose a pipeline to generate Neural Radiance Fields (NeRF) of an object or a scene of a specific class, conditioned on a single input image. This is a challenging task, as training NeRF requires multiple views of the same scene, coupled with corresponding poses, which are hard to obtain. Our method is based on pi-GAN, a generative model for unconditional 3D-aware image synthesis, which maps random latent codes to radiance fields of a class of objects. We jointly optimize (1) the pi-GAN objective to utilize its high-fidelity 3D-aware generation and (2) a carefully designed reconstruction objective. The latter includes an encoder coupled with pi-GAN generator to form an auto-encoder. Unlike previous few-shot NeRF approaches, our pipeline is unsupervised, capable of being trained with independent images without 3D, multi-view, or pose supervision. Applications of our pipeline include 3d avatar generation, object-centric novel view synthesis with a single input image, and 3d-aware super-resolution, to name a few.

The rise of deep neural networks has led to several breakthroughs for semantic segmentation. In spite of this, a model trained on source domain often fails to work properly in new challenging domains, that is directly concerned with the generalization capability of the model. In this paper, we present a novel memory-guided domain generalization method for semantic segmentation based on meta-learning framework. Especially, our method abstracts the conceptual knowledge of semantic classes into categorical memory which is constant beyond the domains. Upon the meta-learning concept, we repeatedly train memory-guided networks and simulate virtual test to 1) learn how to memorize a domain-agnostic and distinct information of classes and 2) offer an externally settled memory as a class-guidance to reduce the ambiguity of representation in the test data of arbitrary unseen domain. To this end, we also propose memory divergence and feature cohesion losses, which encourage to learn memory reading and update processes for category-aware domain generalization. Extensive experiments for semantic segmentation demonstrate the superior generalization capability of our method over state-of-the-art works on various benchmarks.

This paper presents a novel Physically-guided Disentangled Implicit Rendering (PhyDIR) framework for high-fidelity 3D face modeling. The motivation comes from two observations: widely-used graphics renderers yield excessive approximations against photo-realistic imaging, while neural rendering methods are highly entangled to perceive 3D-aware operations. Hence, we learn to disentangle the implicit rendering via explicit physical guidance, meanwhile guarantee the properties of (1) 3D-aware comprehension and (2) high-reality imaging. For the former one, PhyDIR explicitly adopts 3D shading and rasterizing modules to control the renderer, which disentangles the lighting, facial shape and view point from neural reasoning. Specifically, PhyDIR proposes a novel multi-image shading strategy to compensate the monocular limitation, so that the lighting variations are accessible to the neural renderer. For the latter one, PhyDIR learns the face-collection implicit texture to avoid ill-posed intrinsic factorization, then leverages a series of consistency losses to constrain the robustness. With the disentangled method, we make 3D face modeling benefit from both kinds of rendering strategies. Extensive experiments on benchmarks show that PhyDIR obtains superior performance than state-of-the-art explicit/implicit methods, on both geometry/texture modeling.