CVPR 2019 Papers

Deep learning techniques have become the to-go models for most vision-related tasks on 2D images. However, their power has not been fully realised on several tasks in 3D space, e.g., 3D scene understanding. In this work, we jointly address the problems of semantic and instance segmentation of 3D point clouds. Specifically, we develop a multi-task pointwise network that simultaneously performs two tasks: predicting the semantic classes of 3D points and embedding the points into high-dimensional vectors so that points of the same object instance are represented by similar embeddings. We then propose a multi-value conditional random field model to incorporate the semantic and instance labels and formulate the problem of semantic and instance segmentation as jointly optimising labels in the field model. The proposed method is thoroughly evaluated and compared with existing methods on different indoor scene datasets including S3DIS and SceneNN. Experimental results showed the robustness of the proposed joint semantic-instance segmentation scheme over its single components. Our method also achieved state-of-the-art performance on semantic segmentation.

The analysis and modeling of cloth has received a lot of attention in recent years. While recent approaches are focused on woven cloth, we present a novel practical approach for the inference of more complex knitwear structures as well as the respective knitting instructions from only a single image without attached annotations. Knitwear is produced by repeating instances of the same pattern, consisting of grid-like arrangements of a small set of basic stitch types. Our framework addresses the identification and localization of the occurring stitch types, which is challenging due to huge appearance variations. The resulting coarsely localized stitch types are used to infer the underlying grid structure as well as for the extraction of the knitting instruction of pattern repeats, taking into account principles of Gestalt theory. Finally, the derived instructions allow the reproduction of the knitting structures, either as renderings or by actual knitting, as demonstrated in several examples.

Modern computer vision algorithms have brought significant advancement to 3D geometry reconstruction. However, illumination and material reconstruction remain less studied, with current approaches assuming very simplified models for materials and illumination. We introduce Inverse Path Tracing, a novel approach to jointly estimate the material properties of objects and light sources in indoor scenes by using an invertible light transport simulation. We assume a coarse geometry scan, along with corresponding images and camera poses. The key contribution of this work is an accurate and simultaneous retrieval of light sources and physically based material properties (e.g., diffuse reflectance, specular reflectance, roughness, etc.) for the purpose of editing and re-rendering the scene under new conditions. To this end, we introduce a novel optimization method using a differentiable Monte Carlo renderer that computes derivatives with respect to the estimated unknown illumination and material properties. This enables joint optimization for physically correct light transport and material models using a tailored stochastic gradient descent.

Current state-of-the-art instance segmentation methods are not suited for real-time applications like autonomous driving, which require fast execution times at high accuracy. Although the currently dominant proposal-based methods have high accuracy, they are slow and generate masks at a fixed and low resolution. Proposal-free methods, by contrast, can generate masks at high resolution and are often faster, but fail to reach the same accuracy as the proposal-based methods. In this work we propose a new clustering loss function for proposal-free instance segmentation. The loss function pulls the spatial embeddings of pixels belonging to the same instance together and jointly learns an instance-specific clustering bandwidth, maximizing the intersection-over-union of the resulting instance mask. When combined with a fast architecture, the network can perform instance segmentation in real-time while maintaining a high accuracy. We evaluate our method on the challenging Cityscapes benchmark and achieve top results (5% improvement over Mask R-CNN) at more than 10 fps on 2MP images.

In this work, we present CRIB (Continual Recognition Inspired by Babies), a synthetic incremental object learning environment that can produce data that models visual imagery produced by object exploration in early infancy. CRIB is coupled with a new 3D object dataset, Toys-200, that contains 200 unique toy-like object instances, and is also compatible with existing 3D datasets. Through extensive empirical evaluation of state-of-the-art incremental learning algorithms, we find the novel empirical result that repetition can significantly ameliorate the effects of catastrophic forgetting. Furthermore, we find that in certain cases repetition allows for performance approaching that of batch learning algorithms. Finally, we propose an unsupervised incremental learning task with intriguing baseline results.

Convolutional Neural Network based approaches for monocular 3D human pose estimation usually require a large amount of training images with 3D pose annotations. While it is feasible to provide 2D joint annotations for large corpora of in-the-wild images with humans, providing accurate 3D annotations to such in-the-wild corpora is hardly feasible in practice. Most existing 3D labelled data sets are either synthetically created or feature in-studio images. 3D pose estimation algorithms trained on such data often have limited ability to generalize to real world scene diversity. We therefore propose a new deep learning based method for monocular 3D human pose estimation that shows high accuracy and generalizes better to in-the-wild scenes. It has a network architecture that comprises a new disentangled hidden space encoding of explicit 2D and 3D features, and uses supervision by a new learned projection model from predicted 3D pose. Our algorithm can be jointly trained on image data with 3D labels and image data with only 2D labels. It achieves state-of-the-art accuracy on challenging in-the-wild data.

Semantic segmentation requires large amounts of pixel-wise annotations to learn accurate models. In this paper, we present a video prediction-based methodology to scale up training sets by synthesizing new training samples in order to improve the accuracy of semantic segmentation networks. We exploit video prediction models' ability to predict future frames in order to also predict future labels. A joint propagation strategy is also proposed to alleviate mis-alignments in synthesized samples. We demonstrate that training segmentation models on datasets augmented by the synthesized samples leads to significant improvements in accuracy. Furthermore, we introduce a novel boundary label relaxation technique that makes training robust to annotation noise and propagation artifacts along object boundaries. Our proposed methods achieve state-of-the-art mIoUs of 83.5% on Cityscapes and 82.9% on CamVid. Our single model, without model ensembles, achieves 72.8% mIoU on the KITTI semantic segmentation test set, which surpasses the winning entry of the ROB challenge 2018.

Recently, unsupervised exemplar-based image-to-image translation, conditioned on a given exemplar without the paired data, has accomplished substantial advancements. In order to transfer the information from an exemplar to an input image, existing methods often use a normalization technique, e.g., adaptive instance normalization, that controls the channel-wise statistics of an input activation map at a particular layer, such as the mean and the variance. Meanwhile, style transfer approaches similar task to image translation by nature, demonstrated superior performance by using the higher-order statistics such as covariance among channels in representing a style. In detail, it works via whitening (given a zero-mean input feature, transforming its covariance matrix into the identity). followed by coloring (changing the covariance matrix of the whitened feature to those of the style feature). However, applying this approach in image translation is computationally intensive and error-prone due to the expensive time complexity and its non-trivial backpropagation. In response, this paper proposes an end-to-end approach tailored for image translation that efficiently approximates this transformation with our novel regularization methods. We further extend our approach to a group-wise form for memory and time efficiency as well as image quality. Extensive qualitative and quantitative experiments demonstrate that our proposed method is fast, both in training and inference, and highly effective in reflecting the style of an exemplar.

Despite significant recent progress on generative models, controlled generation of images depicting multiple and complex object layouts is still a difficult problem. Among the core challenges are the diversity of appearance a given object may possess and, as a result, exponential set of images consistent with a specified layout. To address these challenges, we propose a novel approach for layout-based image generation; we call it Layout2Im. Given the coarse spatial layout (bounding boxes + object categories), our model can generate a set of realistic images which have the correct objects in the desired locations. The representation of each object is disentangled into a specified/certain part (category) and an unspecified/uncertain part (appearance). The category is encoded using a word embedding and the appearance is distilled into a low-dimensional vector sampled from a normal distribution. Individual object representations are composed together using convolutional LSTM, to obtain an encoding of the complete layout, and then decoded to an image. Several loss terms are introduced to encourage accurate and diverse generation. The proposed Layout2Im model significantly outperforms the previous state of the art, boosting the best reported inception score by 24.66% and 28.57% on the very challenging COCO-Stuff and Visual Genome datasets, respectively. Extensive experiments also demonstrate our method's ability to generate complex and diverse images with multiple objects.

Humans can robustly learn novel visual concepts even when images undergo various deformations and loose certain information. Mimicking the same behavior and synthesizing deformed instances of new concepts may help visual recognition systems perform better one-shot learning, i.e., learning concepts from one or few examples. Our key insight is that, while the deformed images may not be visually realistic, they still maintain critical semantic information and contribute significantly to formulating classifier decision boundaries. Inspired by the recent progress of meta-learning, we combine a meta-learner with an image deformation sub-network that produces additional training examples, and optimize both models in an end-to-end manner. The deformation sub-network learns to deform images by fusing a pair of images --- a probe image that keeps the visual content and a gallery image that diversifies the deformations. We demonstrate results on the widely used one-shot learning benchmarks (miniImageNet and ImageNet 1K Challenge datasets), which significantly outperform state-of-the-art approaches.

Insect pests are one of the main factors affecting agricultural product yield. Accurate recognition of insect pests facilitates timely preventive measures to avoid economic losses. However, the existing datasets for the visual classification task mainly focus on common objects, e.g., flowers and dogs. This limits the application of powerful deep learning technology on specific domains like the agricultural field. In this paper, we collect a large-scale dataset named IP102 for insect pest recognition. Specifically, it contains more than 75, 000 images belonging to 102 categories, which exhibit a natural long-tailed distribution. In addition, we annotate about 19, 000 images with bounding boxes for object detection. The IP102 has a hierarchical taxonomy and the insect pests which mainly affect one specific agricultural product are grouped into the same upperlevel category. Furthermore, we perform several baseline experiments on the IP102 dataset, including handcrafted and deep feature based classification methods. Experimental results show that this dataset has the challenges of interand intra- class variance and data imbalance. We believe our IP102 will facilitate future research on practical insect pest control, fine-grained visual classification, and imbalanced learning fields. We make the dataset and pre-trained models publicly available at https://github.com/xpwu95/IP102.

Video frame interpolation is a long-studied problem in the video processing field. Recently, deep learning approaches have been applied to this problem, showing impressive results on low-resolution benchmarks. However, these methods do not scale-up favorably to high resolutions. Specifically, when the motion exceeds a typical number of pixels, their interpolation quality is degraded. Moreover, their run time renders them impractical for real-time applications. In this paper we propose IM-Net: an interpolated motion neural network. We use an economic structured architecture and end-to-end training with multi-scale tailored losses. In particular, we formulate interpolated motion estimation as classification rather than regression. IM-Net outperforms previous methods by more than 1.3dB (PSNR) on a high resolution version of the recently introduced Vimeo triplet dataset. Moreover, the network runs in less than 33msec on a single GPU for HD resolution.

Generative adversarial networks have achieved great success in unpaired image-to-image translation. Cycle consistency allows modeling the relationship between two distinct domains without paired data. In this paper, we propose an alternative framework, as an extension of latent space interpolation, to consider the intermediate region between two domains during translation. It is based on the fact that in a flat and smooth latent space, there exist many paths that connect two sample points. Properly selecting paths makes it possible to change only certain image attributes, which is useful for generating intermediate images between the two domains. We also show that this framework can be applied to multi-domain and multi-modal translation. Extensive experiments manifest its generality and applicability to various tasks.

We introduce HoloPose, a method for holistic monocular 3D human body reconstruction. We first introduce a part-based model for 3D model parameter regression that allows our method to operate in-the-wild, gracefully handling severe occlusions and large pose variation. We further train a multi-task network comprising 2D, 3D and Dense Pose estimation to drive the 3D reconstruction task. For this we introduce an iterative refinement method that aligns the model-based 3D estimates of 2D/3D joint positions and DensePose with their image-based counterparts delivered by CNNs, achieving both model-based, global consistency and high spatial accuracy thanks to the bottom-up CNN processing. We validate our contributions on challenging benchmarks, showing that our method allows us to get both accurate joint and 3D surface estimates while operating at more than 10fps in-the-wild. More information about our approach, including videos and demos is available at http://arielai.com/holopose.

In radiologists' routine work, one major task is to read a medical image, e.g., a CT scan, find significant lesions, and describe them in the radiology report. In this paper, we study the lesion description or annotation problem. Given a lesion image, our aim is to predict a comprehensive set of relevant labels, such as the lesion's body part, type, and attributes, which may assist downstream fine-grained diagnosis. To address this task, we first design a deep learning module to extract relevant semantic labels from the radiology reports associated with the lesion images. With the images and text-mined labels, we propose a lesion annotation network (LesaNet) based on a multilabel convolutional neural network (CNN) to learn all labels holistically. Hierarchical relations and mutually exclusive relations between the labels are leveraged to improve the label prediction accuracy. The relations are utilized in a label expansion strategy and a reliable hard example mining algorithm. We also attach a simple score propagation layer on LesaNet to enhance recall and explore implicit relation between labels. Multilabel metric learning is combined with classification to enable interpretable prediction. We evaluated LesaNet on the public DeepLesion dataset, which contains over 32K diverse lesion images. Experiments show that LesaNet can precisely annotate the lesions using an ontology of 171 fine-grained labels with an average AUC of 0.9344.

Muscle-based systems have the potential to provide both anatomical accuracy and semantic interpretability as compared to blendshape models; however, a lack of expressivity and differentiability has limited their impact. Thus, we propose modifying a recently developed rather expressive muscle-based system in order to make it fully-differentiable; in fact, our proposed modifications allow this physically robust and anatomically accurate muscle model to conveniently be driven by an underlying blendshape basis. Our formulation is intuitive, natural, as well as monolithically and fully coupled such that one can differentiate the model from end to end, which makes it viable for both optimization and learning-based approaches for a variety of applications. We illustrate this with a number of examples including both shape matching of three-dimensional geometry as as well as the automatic determination of a three-dimensional facial pose from a single two-dimensional RGB image without using markers or depth information.

Single-photon avalanche diodes (SPADs) are an emerging technology with a unique capability of capturing individual photons with high timing precision. SPADs are being used in several active imaging systems (e.g., fluorescence lifetime microscopy and LiDAR), albeit mostly limited to low photon flux settings. We propose passive free-running SPAD (PF-SPAD) imaging, an imaging modality that uses SPADs for capturing 2D intensity images with unprecedented dynamic range under ambient lighting, without any active light source. Our key observation is thatthe precise inter-photon timing measured by a SPAD can be used for estimating scene brightness under ambient lighting conditions, even for very bright scenes. We develop a theoretical model for PF-SPAD imaging, and derive a scene brightness estimator based on the average time of darkness between successive photons detected by a PF-SPAD pixel. Our key insight is that due to the stochastic nature of photon arrivals, this estimator does not suffer from a hard saturation limit. Coupled with high sensitivity at low flux, this enables a PF-SPAD pixel to measure a wide range of scene brightnesses, from very low to very high, thereby achieving extreme dynamic range. We demonstrate an improvement of over 2 orders of magnitude over conventional sensors by imaging scenes spanning a dynamic range of 10^6:1.

This paper presents a hardness-aware deep metric learning (HDML) framework. Most previous deep metric learning methods employ the hard negative mining strategy to alleviate the lack of informative samples for training. However, this mining strategy only utilizes a subset of training data, which may not be enough to characterize the global geometry of the embedding space comprehensively. To address this problem, we perform linear interpolation on embeddings to adaptively manipulate their hard levels and generate corresponding label-preserving synthetics for recycled training, so that information buried in all samples can be fully exploited and the metric is always challenged with proper difficulty. Our method achieves very competitive performance on the widely used CUB-200-2011, Cars196, and Stanford Online Products datasets.

Model quantization is a widely used technique to compress and accelerate deep neural network (DNN) inference. Emergent DNN hardware accelerators begin to support mixed precision (1-8 bits) to further improve the computation efficiency, which raises a great challenge to find the optimal bitwidth for each layer: it requires domain experts to explore the vast design space trading off among accuracy, latency, energy, and model size, which is both time-consuming and sub-optimal. There are plenty of specialized hardware for neural networks, but little research has been done for specialized neural network optimization for a particular hardware architecture. Conventional quantization algorithm ignores the different hardware architectures and quantizes all the layers in a uniform way. In this paper, we introduce the Hardware-Aware Automated Quantization (HAQ) framework which leverages the reinforcement learning to automatically determine the quantization policy, and we take the hardware accelerator's feedback in the design loop. Rather than relying on proxy signals such as FLOPs and model size, we employ a hardware simulator to generate direct feedback signals (latency and energy) to the RL agent. Compared with conventional methods, our framework is fully automated and can specialize the quantization policy for different neural network architectures and hardware architectures. Our framework effectively reduced the latency by 1.4-1.95x and the energy consumption by 1.9x with negligible loss of accuracy compared with the fixed bitwidth (8 bits) quantization. Our framework reveals that the optimal policies on different hardware architectures (i.e., edge and cloud architectures) under different resource constraints (i.e., latency, energy and model size) are drastically different. We interpreted the implication of different quantization policies, which offer insights for both neural network architecture design and hardware architecture design.