CVPR 2020 Papers

Data-driven depth estimation methods struggle with the generalization outside their training scenes due to the immense variability of the real-world scenes. This problem can be partially addressed by utilising synthetically generated images, but closing the synthetic-real domain gap is far from trivial. In this paper, we tackle this issue by using domain invariant defocus blur as direct supervision. We leverage defocus cues by using a permutation invariant convolutional neural network that encourages the network to learn from the differences between images with a different point of focus. Our proposed network uses the defocus map as an intermediate supervisory signal. We are able to train our model completely on synthetic data and directly apply it to a wide range of real-world images. We evaluate our model on synthetic and real datasets, showing compelling generalization results and state-of-the-art depth prediction. The dataset and code are available at https://github.com/dvl-tum/defocus-net.

Applying artificial intelligence techniques in medical imaging is one of the most promising areas in medicine. However, most of the recent success in this area highly relies on large amounts of carefully annotated data, whereas annotating medical images is a costly process. In this paper, we propose a novel method, called FocalMix, which, to the best of our knowledge, is the first to leverage recent advances in semi-supervised learning (SSL) for 3D medical image detection. We conducted extensive experiments on two widely used datasets for lung nodule detection, LUNA16 and NLST. Results show that our proposed SSL methods can achieve a substantial improvement of up to 17.3% over state-of-the-art supervised learning approaches with 400 unlabeled CT scans.

In this paper, we propose a unified method to jointly learn optical flow and stereo matching. Our first intuition is stereo matching can be modeled as a special case of optical flow, and we can leverage 3D geometry behind stereoscopic videos to guide the learning of these two forms of correspondences. We then enroll this knowledge into the state-of-the-art self-supervised learning framework, and train one single network to estimate both flow and stereo. Second, we unveil the bottlenecks in prior self-supervised learning approaches, and propose to create a new set of challenging proxy tasks to boost performance. These two insights yield a single model that achieves the highest accuracy among all existing unsupervised flow and stereo methods on KITTI 2012 and 2015 benchmarks. More remarkably, our self-supervised method even outperforms several state-of-the-art fully supervised methods, including PWC-Net and FlowNet2 on KITTI 2012.

Recent emerged quantization technique (i.e., using low bit-width fixed-point data instead of high bit-width floating-point data) has been applied to inference of deep neural networks for fast and efficient execution. However, directly applying quantization in training can cause significant accuracy loss, thus remaining an open challenge. In this paper, we propose a novel training approach, which applies a layer-wise precision-adaptive quantization in deep neural networks. The new training approach leverages our key insight that the degradation of training accuracy is attributed to the dramatic change of data distribution. Therefore, by keeping the data distribution stable through a layer-wise precision-adaptive quantization, we are able to directly train deep neural networks using low bit-width fixed-point data and achieve guaranteed accuracy, without changing hyper parameters. Experimental results on a wide variety of network architectures (e.g., convolution and recurrent networks) and applications (e.g., image classification, object detection, segmentation and machine translation) show that the proposed approach can train these neural networks with negligible accuracy losses (-1.40%-1.3%, 0.02% on average), and speed up training by 252% on a state-of-the-art Intel CPU.

Cross-modal retrieval between videos and texts has attracted growing attentions due to the rapid emergence of videos on the web. The current dominant approach is to learn a joint embedding space to measure cross-modal similarities. However, simple embeddings are insufficient to represent complicated visual and textual details, such as scenes, objects, actions and their compositions. To improve fine-grained video-text retrieval, we propose a Hierarchical Graph Reasoning (HGR) model, which decomposes video-text matching into global-to-local levels. The model disentangles text into a hierarchical semantic graph including three levels of events, actions, entities, and generates hierarchical textual embeddings via attention-based graph reasoning. Different levels of texts can guide the learning of diverse and hierarchical video representations for cross-modal matching to capture both global and local details. Experimental results on three video-text datasets demonstrate the advantages of our model. Such hierarchical decomposition also enables better generalization across datasets and improves the ability to distinguish fine-grained semantic differences. Code will be released at https://github.com/cshizhe/hgr_v2t.

Existing image-to-image transformation approaches primarily focus on synthesizing visually pleasing data. Generating images with correct identity labels is challenging yet much less explored. It is even more challenging to deal with image transformation tasks with large deformation in poses, viewpoints, or scales while preserving the identity, such as face rotation and object viewpoint morphing. In this paper, we aim at transforming an image with a fine-grained category to synthesize new images that preserve the identity of the input image, which can thereby benefit the subsequent fine-grained image recognition and few-shot learning tasks. The generated images, transformed with large geometric deformation, do not necessarily need to be of high visual quality but are required to maintain as much identity information as possible. To this end, we adopt a model based on generative adversarial networks to disentangle the identity related and unrelated factors of an image. In order to preserve the fine-grained contextual details of the input image during the deformable transformation, a constrained nonalignment connection method is proposed to construct learnable highways between intermediate convolution blocks in the generator. Moreover, an adaptive identity modulation mechanism is proposed to transfer the identity information into the output image effectively. Extensive experiments on the CompCars and Multi-PIE datasets demonstrate that our model preserves the identity of the generated images much better than the state-of-the-art image-to-image transformation models, and as a result significantly boosts the visual recognition performance in fine-grained few-shot learning.

We address the problem of fine-grained generalized zero-shot recognition of visually similar classes without training images for some classes. We propose a dense attribute-based attention mechanism that for each attribute focuses on the most relevant image regions, obtaining attribute-based features. Instead of aligning a global feature vector of an image with its associated class semantic vector, we propose an attribute embedding technique that aligns each attribute-based feature with its attribute semantic vector. Hence, we compute a vector of attribute scores, for the presence of each attribute in an image, whose similarity with the true class semantic vector is maximized. Moreover, we adjust each attribute score using an attention mechanism over attributes to better capture the discriminative power of different attributes. To tackle the challenge of bias towards seen classes during testing, we propose a new self-calibration loss that adjusts the probability of unseen classes to account for the training bias. We conduct experiments on three popular datasets of CUB, SUN and AWA2 as well as the large-scale DeepFashion dataset, showing that our model significantly improves the state of the art.

This paper proposes a new learning paradigm called filter grafting, which aims to improve the representation capability of Deep Neural Networks (DNNs). The motivation is that DNNs have unimportant (invalid) filters (e.g., l1 norm close to 0). These filters limit the potential of DNNs since they are identified as having little effect on the network. While filter pruning removes these invalid filters for efficiency consideration, filter grafting re-activates them from an accuracy boosting perspective. The activation is processed by grafting external information (weights) into invalid filters. To better perform the grafting process, we develop an entropy-based criterion to measure the information of filters and an adaptive weighting strategy for balancing the grafted information among networks. After the grafting operation, the network has very few invalid filters compared with its untouched state, enpowering the model with more representation capacity. We also perform extensive experiments on the classification and recognition tasks to show the superiority of our method. For example, the grafted MobileNetV2 outperforms the non-grafted MobileNetV2 by about 7 percent on CIFAR-100 dataset.

Difficulty in collecting and annotating large-scale video data raises a growing interest in learning models which can recognize novel classes with only a few training examples. In this paper, we propose the Ordered Temporal Alignment Module (OTAM), a novel few-shot learning framework that can learn to classify a previously unseen video. While most previous work neglects long-term temporal ordering information, our proposed model explicitly leverages the temporal ordering information in video data through ordered temporal alignment. This leads to strong data-efficiency for few-shot learning. In concrete, our proposed pipeline learns a deep distance measurement of the query video with respect to novel class proxies over its alignment path. We adopt an episode-based training scheme and directly optimize the few-shot learning objective. We evaluate OTAM on two challenging real-world datasets, Kinetics and Something-Something-V2, and show that our model leads to significant improvement of few-shot video classification over a wide range of competitive baselines and outperforms state-of-the-art benchmarks by a large margin.

The problem of open-set recognition is considered. While previous approaches only consider this problem in the context of large-scale classifier training, we seek a unified solution for this and the low-shot classification setting. It is argued that the classic softmax classifier is a poor solution for open-set recognition, since it tends to overfit on the training classes. Randomization is then proposed as a solution to this problem. This suggests the use of meta-learning techniques, commonly used for few-shot classification, for the solution of open-set recognition. A new oPen sEt mEta LEaRning (PEELER) algorithm is then introduced. This combines the random selection of a set of novel classes per episode, a loss that maximizes the posterior entropy for examples of those classes, and a new metric learning formulation based on the Mahalanobis distance. Experimental results show that PEELER achieves state of the art open set recognition performance for both few-shot and large-scale recognition. On CIFAR and miniImageNet, it achieves substantial gains in seen/unseen class detection AUROC for a given seen-class classification accuracy.

Conventional methods for object detection typically require a substantial amount of training data and preparing such high-quality training data is very labor-intensive. In this paper, we propose a novel few-shot object detection network that aims at detecting objects of unseen categories with only a few annotated examples. Central to our method are our Attention-RPN, Multi-Relation Detector and Contrastive Training strategy, which exploit the similarity between the few shot support set and query set to detect novel objects while suppressing false detection in the background. To train our network, we contribute a new dataset that contains 1000 categories of various objects with high-quality annotations. To the best of our knowledge, this is one of the first datasets specifically designed for few-shot object detection. Once our few-shot network is trained, it can detect objects of unseen categories without further training or fine-tuning. Our method is general and has a wide range of potential applications. We produce a new state-of-the-art performance on different datasets in the few-shot setting. The dataset link is https://github.com/fanq15/Few-Shot-Object-Detection-Dataset.

Learning with limited data is a key challenge for visual recognition. Many few-shot learning methods address this challenge by learning an instance embedding function from seen classes and apply the function to instances from unseen classes with limited labels. This style of transfer learning is task-agnostic: the embedding function is not learned optimally discriminative with respect to the unseen classes, where discerning among them leads to the target task. In this paper, we propose a novel approach to adapt the instance embeddings to the target classification task with a set-to-set function, yielding embeddings that are task-specific and are discriminative. We empirically investigated various instantiations of such set-to-set functions and observed the Transformer is most effective --- as it naturally satisfies key properties of our desired model. We denote this model as FEAT (few-shot embedding adaptation w/ Transformer) and validate it on both the standard few-shot classification benchmark and four extended few-shot learning settings with essential use cases, i.e., cross-domain, transductive, generalized few-shot learning, and low-shot learning. It archived consistent improvements over baseline models as well as previous methods, and established the new state-of-the-art results on two benchmarks.

Recently, deep neural networks are introduced as supervised discriminative models for the learning of 3D point cloud segmentation. Most previous supervised methods require a large number of training data with human annotation part labels to guide the training process to ensure the model's generalization abilities on test data. In comparison, we propose a novel 3D shape segmentation method that requires few labeled data for training. Given an input 3D shape, the training of our model starts with identifying a similar 3D shape with part annotations from a mini-pool of shape templates (e.g. 10 shapes). With the selected template shape, a novel Coherent Point Transformer is proposed to fully leverage the power of a deep neural network to smoothly morph the template shape towards the input shape. Then, based on the transformed template shapes with part labels, a newly proposed Part-specific Density Estimator is developed to learn a continuous part-specific probability distribution function on the entire 3D space with a batch consistency regularization term. With the learned part-specific probability distribution, our model is able to predict the part labels of a new input 3D shape in an end-to-end manner. We demonstrate that our proposed method can achieve remarkable segmentation results on the ShapeNet dataset with few shots, compared to previous supervised learning approaches.

Deep neural network compression techniques such as pruning and weight tensor decomposition usually require fine-tuning to recover the prediction accuracy when the compression ratio is high. However, conventional fine-tuning suffers from the requirement of a large training set and the time-consuming training procedure. This paper proposes a novel solution for knowledge distillation from label-free few samples to realize both data efficiency and training/processing efficiency. We treat the original network as "teacher-net" and the compressed network as "student-net". A 1x1 convolution layer is added at the end of each layer block of the student-net, and we fit the block-level outputs of the student-net to the teacher-net by estimating the parameters of the added layers. We prove that the added layer can be merged without adding extra parameters and computation cost during inference. Experiments on multiple datasets and network architectures verify the method's effectiveness on student-nets obtained by various network pruning and weight decomposition methods. Our method can recover student-net's accuracy to the same level as conventional fine-tuning methods in minutes while using only 1% label-free data of the full training data.

Video interpolation aims to synthesize non-existent frames between two consecutive frames. Although existing optical flow based methods have achieved promising results, they still face great challenges in dealing with the interpolation of complicated dynamic scenes, which include occlusion, blur or abrupt brightness change. This is mainly because these cases may break the basic assumptions of the optical flow estimation (i.e. smoothness, consistency). In this work, we devised a novel structure-to-texture generation framework which splits the video interpolation task into two stages: structure-guided interpolation and texture refinement. In the first stage, deep structure-aware features are employed to predict feature flows from two consecutive frames to their intermediate result, and further generate the structure image of the intermediate frame. In the second stage, based on the generated coarse result, a Frame Texture Compensator is trained to fill in detailed textures. To the best of our knowledge, this is the first work that attempts to directly generate the intermediate frame through blending deep features. Experiments on both the benchmark datasets and challenging occlusion cases demonstrate the superiority of the proposed framework over the state-of-the-art methods. Codes are available on https://github.com/CM-BF/FeatureFlow.

We present a fast feature-metric point cloud registration framework, which enforces the optimisation of registration by minimising a feature-metric projection error without correspondences. The advantage of the feature-metric projection error is robust to noise, outliers and density difference in contrast to the geometric projection error. Besides, minimising the feature-metric projection error does not need to search the correspondences so that the optimisation speed is fast. The principle behind the proposed method is that the feature difference is smallest if point clouds are aligned very well. We train the proposed method in a semi-supervised or unsupervised approach, which requires limited or no registration label data. Experiments demonstrate our method obtains higher accuracy and robustness than the state-of-the-art methods. Besides, experimental results show that the proposed method can handle significant noise and density difference, and solve both same-source and cross-source point cloud registration.

Almost all previous deep learning-based multi-view stereo (MVS) approaches focus on improving reconstruction quality. Besides quality, efficiency is also a desirable feature for MVS in real scenarios. Towards this end, this paper presents a Fast-MVSNet, a novel sparse-to-dense coarse-to-fine framework, for fast and accurate depth estimation in MVS. Specifically, in our Fast-MVSNet, we first construct a sparse cost volume for learning a sparse and high-resolution depth map. Then we leverage a small-scale convolutional neural network to encode the depth dependencies for pixels within a local region to densify the sparse high-resolution depth map. At last, a simple but efficient Gauss-Newton layer is proposed to further optimize the depth map. On one hand, the high-resolution depth map, the data-adaptive propagation method and the Gauss-Newton layer jointly guarantee the effectiveness of our method. On the other hand, all modules in our Fast-MVSNet are lightweight and thus guarantee the efficiency of our approach. Besides, our approach is also memory-friendly because of the sparse depth representation. Extensive experimental results show that our method is 5 times and 14 times faster than Point-MVSNet and R-MVSNet, respectively, while achieving comparable or even better results on the challenging Tanks and Temples dataset as well as the DTU dataset. Code is available at https://github.com/svip-lab/FastMVSNet.

The reconstruction of shredded documents consists in arranging the pieces of paper (shreds) in order to reassemble the original aspect of such documents. This task is particularly relevant for supporting forensic investigation as documents may contain criminal evidence. As an alternative to the laborious and time-consuming manual process, several researchers have been investigating ways to perform automatic digital reconstruction. A central problem in automatic reconstruction of shredded documents is the pairwise compatibility evaluation of the shreds, notably for binary text documents. In this context, deep learning has enabled great progress for accurate reconstructions in the domain of mechanically-shredded documents. A sensitive issue, however, is that current deep model solutions require an inference whenever a pair of shreds has to be evaluated. This work proposes a scalable deep learning approach for measuring pairwise compatibility in which the number of inferences scales linearly (rather than quadratically) with the number of shreds. Instead of predicting compatibility directly, deep models are leveraged to asymmetrically project the raw shred content onto a common metric space in which distance is proportional to the compatibility. Experimental results show that our method has accuracy comparable to the state-of-the-art with a speed-up of about 22 times for a test instance with 505 shreds (20 mixed shredded-pages from different documents).

Significant progress has been made in Video Object Segmentation (VOS), the video object tracking task in its finest level. While the VOS task can be naturally decoupled into image semantic segmentation and video object tracking, significantly much more research effort has been made in segmentation than tracking. In this paper, we introduce "tracking-by-detection" into VOS which can coherently integrates segmentation into tracking, by proposing a new temporal aggregation network and a novel dynamic time-evolving template matching mechanism to achieve significantly improved performance. Notably, our method is entirely online and thus suitable for one-shot learning, and our end-to-end trainable model allows multiple object segmentation in one forward pass. We achieve new state-of-the-art performance on the DAVIS benchmark without complicated bells and whistles in both speed and accuracy, with a speed of 0.14 second per frame and J &F measure of 75.9% respectively.