Fast Camera Image Denoising on Mobile GPUs with Deep Learning, Mobile AI 2021 Challenge: Report

Image denoising is one of the most critical problems in mobile photo processing. While many solutions have been proposed for this task, they are usually working with synthetic data and are too computationally expensive to run on mobile devices. To address this problem, we introduce the first Mobile AI challenge, where the target is to develop an end-to-end deep learning-based image denoising solution that can demonstrate high efficiency on smartphone GPUs. For this, the participants were provided with a novel large-scale dataset consisting of noisy-clean image pairs captured in the wild. The runtime of all models was evaluated on the Samsung Exynos 2100 chipset with a powerful Mali GPU capable of accelerating floating-point and quantized neural networks. The proposed solutions are fully compatible with any mobile GPU and are capable of processing 480p resolution images under 40-80 ms while achieving high fidelity results. A detailed description of all models developed in the challenge is provided in this paper.

Practical Wide-Angle Portraits Correction with Deep Structured Models

Wide-angle portraits often enjoy expanded views. However, they contain perspective distortions, especially noticeable when capturing group portrait photos, where the background is skewed and faces are stretched. This paper introduces the first deep learning based approach to remove such artifacts from freely-shot photos. Specifically, given a wide-angle portrait as input, we build a cascaded network consisting of a LineNet, a ShapeNet, and a transition module (TM), which corrects perspective distortions on the background, adapts to the stereographic projection on facial regions, and achieves smooth transitions between these two projections, accordingly. To train our network, we build the first perspective portrait dataset with a large diversity in identities, scenes and camera modules. For the quantitative evaluation, we introduce two novel metrics, line consistency and face congruence. Compared to the previous state-of-the-art approach, our method does not require camera distortion parameters. We demonstrate that our approach significantly outperforms the previous state-of-the-art approach both qualitatively and quantitatively.

ASFlow: Unsupervised Optical Flow Learning with Adaptive Pyramid Sampling

We present an unsupervised optical flow estimation method by proposing an adaptive pyramid sampling in the deep pyramid network. Specifically, in the pyramid downsampling, we propose an Content Aware Pooling (CAP) module, which promotes local feature gathering by avoiding cross region pooling, so that the learned features become more representative. In the pyramid upsampling, we propose an Adaptive Flow Upsampling (AFU) module, where cross edge interpolation can be avoided, producing sharp motion boundaries. Equipped with these two modules, our method achieves the best performance for unsupervised optical flow estimation on multiple leading benchmarks, including MPI-SIntel, KITTI 2012 and KITTI 2015. Particuarlly, we achieve EPE=1.5 on KITTI 2012 and F1=9.67% KITTI 2015, which outperform the previous state-of-the-art methods by 16.7% and 13.1%, respectively.

Motion Basis Learning for Unsupervised Deep Homography Estimation with Subspace Projection

In this paper, we introduce a new framework for unsupervised deep homography estimation. Our contributions are 3 folds. First, unlike previous methods that regress 4 offsets for a homography, we propose a homography flow representation, which can be estimated by a weighted sum of 8 pre-defined homography flow bases. Second, considering a homography contains 8 Degree-of-Freedoms (DOFs) that is much less than the rank of the network features, we propose a Low Rank Representation (LRR) block that reduces the feature rank, so that features corresponding to the dominant motions are retained while others are rejected. Last, we propose a Feature Identity Loss (FIL) to enforce the learned image feature warp-equivariant, meaning that the result should be identical if the order of warp operation and feature extraction is swapped. With this constraint, the unsupervised optimization is achieved more effectively and more stable features are learned. Extensive experiments are conducted to demonstrate the effectiveness of all the newly proposed components, and results show our approach outperforms the state-of-the-art on the homography benchmark datasets both qualitatively and quantitatively.

D2C-SR: A Divergence to Convergence Approach for Image Super-Resolution

In this paper, we present D2C-SR, a novel framework for the task of image super-resolution(SR). As an ill-posed problem, the key challenge for super-resolution related tasks is there can be multiple predictions for a given low-resolution input. Most classical methods and early deep learning based approaches ignored this fundamental fact and modeled this problem as a deterministic processing which often lead to unsatisfactory results. Inspired by recent works like SRFlow, we tackle this problem in a semi-probabilistic manner and propose a two-stage pipeline: a divergence stage is used to learn the distribution of underlying high-resolution outputs in a discrete form, and a convergence stage is followed to fuse the learned predictions into a final output. More specifically, we propose a tree-based structure deep network, where each branch is designed to learn a possible high-resolution prediction. At the divergence stage, each branch is trained separately to fit ground truth, and a triple loss is used to enforce the outputs from different branches divergent. Subsequently, we add a fuse module to combine the multiple predictions as the outputs from the first stage can be sub-optimal. The fuse module can be trained to converge w.r.t the final high-resolution image in an end-to-end manner. We conduct evaluations on several benchmarks, including a new proposed dataset with 8x upscaling factor. Our experiments demonstrate that D2C-SR can achieve state-of-the-art performance on PSNR and SSIM, with a significantly less computational cost.

GyroFlow: Gyroscope-Guided Unsupervised Optical Flow Learning

Existing optical flow methods are erroneous in challenging scenes, such as fog, rain, and night because the basic optical flow assumptions such as brightness and gradient constancy are broken. To address this problem, we present an unsupervised learning approach that fuses gyroscope into optical flow learning. Specifically, we first convert gyroscope readings into motion fields named gyro field. Then, we design a self-guided fusion module to fuse the background motion extracted from the gyro field with the optical flow and guide the network to focus on motion details. To the best of our knowledge, this is the first deep learning-based framework that fuses gyroscope data and image content for optical flow learning. To validate our method, we propose a new dataset that covers regular and challenging scenes. Experiments show that our method outperforms the state-of-art methods in both regular and challenging scenes.

Holistic 3D Scene Understanding from a Single Image with Implicit Representation

We present a new pipeline for holistic 3D scene understanding from a single image, which could predict object shape, object pose, and scene layout. As it is a highly ill-posed problem, existing methods usually suffer from inaccurate estimation of both shapes and layout especially for the cluttered scene due to the heavy occlusion between objects. We propose to utilize the latest deep implicit representation to solve this challenge. We not only propose an image-based local structured implicit network to improve the object shape estimation, but also refine 3D object pose and scene layout via a novel implicit scene graph neural network that exploits the implicit local object features. A novel physical violation loss is also proposed to avoid incorrect context between objects. Extensive experiments demonstrate that our method outperforms the state-of-the-art methods in terms of object shape, scene layout estimation, and 3D object detection.

OMNet: Learning Overlapping Mask for Partial-to-Partial Point Cloud Registration

Point cloud registration is a key task in many computational fields. Previous correspondence matching based methods require the point clouds to have distinctive geometric structures to fit a 3D rigid transformation according to point-wise sparse feature matches. However, the accuracy of transformation heavily relies on the quality of extracted features, which are prone to errors with respect partiality and noise of the inputs. In addition, they can not utilize the geometric knowledge of all regions. On the other hand, previous global feature based deep learning approaches can utilize the entire point cloud for the registration, however they ignore the negative effect of non-overlapping points when aggregating global feature from point-wise features. In this paper, we present OMNet, a global feature based iterative network for partial-to-partial point cloud registration. We learn masks in a coarse-to-fine manner to reject non-overlapping regions, which converting the partial-to-partial registration to the registration of the same shapes. Moreover, the data used in previous works are only sampled once from CAD models for each object, resulting the same point cloud for the source and the reference. We propose a more practical manner for data generation, where a CAD model is sampled twice for the source and the reference point clouds, avoiding over-fitting issues that commonly exist previously. Experimental results show that our approach achieves state-of-the-art performance compared to traditional and deep learning methods.

UPHDR-GAN: Generative Adversarial Network for High Dynamic Range Imaging with Unpaired Data

The paper proposes a method to effectively fuse multi-exposure inputs and generates high-quality high dynamic range (HDR) images with unpaired datasets. Deep learning-based HDR image generation methods rely heavily on paired datasets. The ground truth provides information for the network getting HDR images without ghosting. Datasets without ground truth are hard to apply to train deep neural networks. Recently, Generative Adversarial Networks (GAN) have demonstrated their potentials of translating images from source domain X to target domain Y in the absence of paired examples. In this paper, we propose a GAN-based network for solving such problems while generating enjoyable HDR results, named UPHDR-GAN. The proposed method relaxes the constraint of paired dataset and learns the mapping from LDR domain to HDR domain. Although the pair data are missing, UPHDR-GAN can properly handle the ghosting artifacts caused by moving objects or misalignments with the help of modified GAN loss, improved discriminator network and useful initialization phase. The proposed method preserves the details of important regions and improves the total image perceptual quality. Qualitative and quantitative comparisons against other methods demonstrated the superiority of our method.