Recovering the 3D shape of an object from single or multiple images with deep neural networks has been attracting increasing attention in the past few years. Mainstream works (e.g. 3D-R2N2) use recurrent neural networks (RNNs) to sequentially fuse feature maps of input images. However, RNN-based approaches are unable to produce consistent reconstruction results when given the same input images with different orders. Moreover, RNNs may forget important features from early input images due to long-term memory loss. To address these issues, we propose a novel framework for single-view and multi-view 3D object reconstruction, named Pix2Vox++. By using a well-designed encoder-decoder, it generates a coarse 3D volume from each input image. A multi-scale context-aware fusion module is then introduced to adaptively select high-quality reconstructions for different parts from all coarse 3D volumes to obtain a fused 3D volume. To further correct the wrongly recovered parts in the fused 3D volume, a refiner is adopted to generate the final output. Experimental results on the ShapeNet, Pix3D, and Things3D benchmarks show that Pix2Vox++ performs favorably against state-of-the-art methods in terms of both accuracy and efficiency.
Applications that involve supervised training require paired images. Researchers of single image super-resolution (SISR) create such images by artificially generating blurry input images from the corresponding ground truth. Similarly we can create paired images with the canny edge. We propose Generator From Edges (GFE) [Figure 2]. Our aim is to determine the best architecture for GFE, along with reviews of perceptual loss [1, 2]. To this end, we conducted three experiments. First, we explored the effects of the adversarial loss often used in SISR. In particular, we uncovered that it is not an essential component to form a perceptual loss. Eliminating adversarial loss will lead to a more effective architecture from the perspective of hardware resource. It also means that considerations for the problems pertaining to generative adversarial network (GAN) [3], such as mode collapse, are not necessary. Second, we reexamined VGG loss and found that the mid-layers yield the best results. By extracting the full potential of VGG loss, the overall performance of perceptual loss improves significantly. Third, based on the findings of the first two experiments, we reevaluated the dense network to construct GFE. Using GFE as an intermediate process, reconstructing a facial image from a pencil sketch can become an easy task.
To improve the performance of most neuroimiage analysis pipelines, brain extraction is used as a fundamental first step in the image processing. But in the case of fetal brain development, there is a need for a reliable US-specific tool. In this work we propose a fully automated 3D CNN approach to fetal brain extraction from 3D US clinical volumes with minimal preprocessing. Our method accurately and reliably extracts the brain regardless of the large data variation inherent in this imaging modality. It also performs consistently throughout a gestational age range between 14 and 31 weeks, regardless of the pose variation of the subject, the scale, and even partial feature-obstruction in the image, outperforming all current alternatives.
Denoisers trained with synthetic data often fail to cope with the diversity of unknown noises, giving way to methods that can adapt to existing noise without knowing its ground truth. Previous image-based method leads to noise overfitting if directly applied to video denoisers, and has inadequate temporal information management especially in terms of occlusion and lighting variation, which considerably hinders its denoising performance. In this paper, we propose a general framework for video denoising networks that successfully addresses these challenges. A novel twin sampler assembles training data by decoupling inputs from targets without altering semantics, which not only effectively solves the noise overfitting problem, but also generates better occlusion masks efficiently by checking optical flow consistency. An online denoising scheme and a warping loss regularizer are employed for better temporal alignment. Lighting variation is quantified based on the local similarity of aligned frames. Our method consistently outperforms the prior art by 0.6-3.2dB PSNR on multiple noises, datasets and network architectures. State-of-the-art results on reducing model-blind video noises are achieved. Extensive ablation studies are conducted to demonstrate the significance of each technical components.
The widely adopted sequential variant of Non Maximum Suppression (or Greedy-NMS) is a crucial module for object-detection pipelines. Unfortunately, for the region proposal stage of two/multi-stage detectors, NMS is turning out to be a latency bottleneck due to its sequential nature. In this article, we carefully profile Greedy-NMS iterations to find that a major chunk of computation is wasted in comparing proposals that are already far-away and have a small chance of suppressing each other. We address this issue by comparing only those proposals that are generated from nearby anchors. The translation-invariant property of the anchor lattice affords generation of a lookup table, which provides an efficient access to nearby proposals, during NMS. This leads to an Accelerated NMS algorithm which leverages Spatially Aware Priors, or ASAP-NMS, and improves the latency of the NMS step from 13.6ms to 1.2 ms on a CPU without sacrificing the accuracy of a state-of-the-art two-stage detector on COCO and VOC datasets. Importantly, ASAP-NMS is agnostic to image resolution and can be used as a simple drop-in module during inference. Using ASAP-NMS at run-time only, we obtain an mAP of 44.2\%@25Hz on the COCO dataset with a V100 GPU.
The primary issue in inverse halftoning is removing noisy dots on flat areas and restoring image structures (e.g., lines, patterns) on textured areas. Hence, a new structure-aware deep convolutional neural network that incorporates two subnetworks is proposed in this paper. One subnetwork is for image structure prediction while the other is for continuous-tone image reconstruction. First, to predict image structures, patch pairs comprising continuous-tone patches and the corresponding halftoned patches generated through digital halftoning are trained. Subsequently, gradient patches are generated by convolving gradient filters with the continuous-tone patches. The subnetwork for the image structure prediction is trained using the mini-batch gradient descent algorithm given the halftoned patches and gradient patches, which are fed into the input and loss layers of the subnetwork, respectively. Next, the predicted map including the image structures is stacked on the top of the input halftoned image through a fusion layer and fed into the image reconstruction subnetwork such that the entire network is trained adaptively to the image structures. The experimental results confirm that the proposed structure-aware network can remove noisy dot-patterns well on flat areas and restore details clearly on textured areas. Furthermore, it is demonstrated that the proposed method surpasses the conventional state-of-the-art methods based on deep convolutional neural networks and locally learned dictionaries.
High-dimensional latent representations learned by neural network classifiers are notoriously hard to interpret. Especially in medical applications, model developers and domain experts desire a better understanding of how these latent representations relate to the resulting classification performance. We present a framework for retraining classifiers by backpropagating manual changes made to low-dimensional embeddings of the latent space. This means that our technique allows the practitioner to control the latent decision space in an intuitive way. Our approach is based on parametric approximations of non-linear embedding techniques such as t-distributed stochastic neighbourhood embedding. Using this approach, it is possible to manually shape and declutter the latent space of image classifiers in order to better match the expectations of domain experts or to fulfil specific requirements of classification tasks. For instance, the performance for specific class pairs can be enhanced by manually separating the class clusters in the embedding, without significantly affecting the overall performance of the other classes. We evaluate our technique on a real-world scenario in fetal ultrasound imaging.
In this paper, we study two important problems in the automated design of neural networks -- Hyper-parameter Optimization (HPO), and Neural Architecture Search (NAS) -- through the lens of sparse recovery methods. In the first part of this paper, we establish a novel connection between HPO and structured sparse recovery. In particular, we show that a special encoding of the hyperparameter space enables a natural group-sparse recovery formulation, which when coupled with HyperBand (a multi-armed bandit strategy), leads to improvement over existing hyperparameter optimization methods. Experimental results on image datasets such as CIFAR-10 confirm the benefits of our approach. In the second part of this paper, we establish a connection between NAS and structured sparse recovery. Building upon ``one-shot'' approaches in NAS, we propose a novel algorithm that we call CoNAS by merging ideas from one-shot approaches with a techniques for learning low-degree sparse Boolean polynomials. We provide theoretical analysis on the number of validation error measurements. Finally, we validate our approach on several datasets and discover novel architectures hitherto unreported, achieving competitive (or better) results in both performance and search time compared to the existing NAS approaches.
Set-based person re-identification (SReID) is a matching problem that aims to verify whether two sets are of the same identity (ID). Existing SReID models typically generate a feature representation per image and aggregate them to represent the set as a single embedding. However, they can easily be perturbed by noises--perceptually/semantically low quality images--which are inevitable due to imperfect tracking/detection systems, or overfit to trivial images. In this work, we present a novel and simple solution to this problem based on ID-aware quality that measures the perceptual and semantic quality of images guided by their ID information. Specifically, we propose an ID-aware Embedding that consists of two key components: (1) Feature learning attention that aims to learn robust image embeddings by focusing on 'medium' hard images. This way it can prevent overfitting to trivial images, and alleviate the influence of outliers. (2) Feature fusion attention is to fuse image embeddings in the set to obtain the set-level embedding. It ignores noisy information and pays more attention to discriminative images to aggregate more discriminative information. Experimental results on four datasets show that our method outperforms state-of-the-art approaches despite the simplicity of our approach.
Generative Adversarial Networks (GANs) have been widely used for generating photo-realistic images. A variant of GANs called super-resolution GAN (SRGAN) has already been used successfully for image super-resolution where low resolution images can be upsampled to a $4\times$ larger image that is perceptually more realistic. However, when such generative models are used for data describing physical processes, there are additional known constraints that models must satisfy including governing equations and boundary conditions. In general, these constraints may not be obeyed by the generated data. In this work, we develop physics-based methods for generative enrichment of turbulence. We incorporate a physics-informed learning approach by a modification to the loss function to minimize the residuals of the governing equations for the generated data. We have analyzed two trained physics-informed models: a supervised model based on convolutional neural networks (CNN) and a generative model based on SRGAN: Turbulence Enrichment GAN (TEGAN), and show that they both outperform simple bicubic interpolation in turbulence enrichment. We have also shown that using the physics-informed learning can also significantly improve the model's ability in generating data that satisfies the physical governing equations. Finally, we compare the enriched data from TEGAN to show that it is able to recover statistical metrics of the flow field including energy metrics and well as inter-scale energy dynamics and flow morphology.