A Coarse-to-Fine Framework for Learned Color Enhancement with Non-Local Attention

Automatic color enhancement is aimed to adaptively adjust photos to expected styles and tones. For current learned methods in this field, global harmonious perception and local details are hard to be well-considered in a single model simultaneously. To address this problem, we propose a coarse-to-fine framework with non-local attention for color enhancement in this paper. Within our framework, we propose to divide enhancement process into channel-wise enhancement and pixel-wise refinement performed by two cascaded Convolutional Neural Networks (CNNs). In channel-wise enhancement, our model predicts a global linear mapping for RGB channels of input images to perform global style adjustment. In pixel-wise refinement, we learn a refining mapping using residual learning for local adjustment. Further, we adopt a non-local attention block to capture the long-range dependencies from global information for subsequent fine-grained local refinement. We evaluate our proposed framework on the commonly using benchmark and conduct sufficient experiments to demonstrate each technical component within it.

Stereoscopic Omnidirectional Image Quality Assessment Based on Predictive Coding Theory

Objective quality assessment of stereoscopic omnidirectional images is a challenging problem since it is influenced by multiple aspects such as projection deformation, field of view (FoV) range, binocular vision, visual comfort, etc. Existing studies show that classic 2D or 3D image quality assessment (IQA) metrics are not able to perform well for stereoscopic omnidirectional images. However, very few research works have focused on evaluating the perceptual visual quality of omnidirectional images, especially for stereoscopic omnidirectional images. In this paper, based on the predictive coding theory of the human vision system (HVS), we propose a stereoscopic omnidirectional image quality evaluator (SOIQE) to cope with the characteristics of 3D 360-degree images. Two modules are involved in SOIQE: predictive coding theory based binocular rivalry module and multi-view fusion module. In the binocular rivalry module, we introduce predictive coding theory to simulate the competition between high-level patterns and calculate the similarity and rivalry dominance to obtain the quality scores of viewport images. Moreover, we develop the multi-view fusion module to aggregate the quality scores of viewport images with the help of both content weight and location weight. The proposed SOIQE is a parametric model without necessary of regression learning, which ensures its interpretability and generalization performance. Experimental results on our published stereoscopic omnidirectional image quality assessment database (SOLID) demonstrate that our proposed SOIQE method outperforms state-of-the-art metrics. Furthermore, we also verify the effectiveness of each proposed module on both public stereoscopic image datasets and panoramic image datasets.

ZstGAN: An Adversarial Approach for Unsupervised Zero-Shot Image-to-Image Translation

Image-to-image translation models have shown remarkable ability on transferring images among different domains. Most of existing work follows the setting that the source domain and target domain keep the same at training and inference phases, which cannot be generalized to the scenarios for translating an image from an unseen domain to an another unseen domain. In this work, we propose the Unsupervised Zero-Shot Image-to-image Translation (UZSIT) problem, which aims to learn a model that can transfer translation knowledge from seen domains to unseen domains. Accordingly, we propose a framework called ZstGAN: By introducing an adversarial training scheme, ZstGAN learns to model each domain with domain-specific feature distribution that is semantically consistent on vision and attribute modalities. Then the domain-invariant features are disentangled with an shared encoder for image generation. We carry out extensive experiments on CUB and FLO datasets, and the results demonstrate the effectiveness of proposed method on UZSIT task. Moreover, ZstGAN shows significant accuracy improvements over state-of-the-art zero-shot learning methods on CUB and FLO.

Learning to Transfer: Unsupervised Meta Domain Translation

Unsupervised domain translation has recently achieved impressive performance with rapidly developed generative adversarial network (GAN) and availability of sufficient training data. However, existing domain translation frameworks form in a disposable way where the learning experiences are ignored. In this work, we take this research direction toward unsupervised meta domain translation problem. We propose a meta translation model called MT-GAN to find parameter initialization of a conditional GAN, which can quickly adapt for a new domain translation task with limited training samples. In the meta-training procedure, MT-GAN is explicitly fine-tuned with a primary translation task and a synthesized dual translation task. Then we design a meta-optimization objective to require the fine-tuned MT-GAN to produce good generalization performance. We demonstrate effectiveness of our model on ten diverse two-domain translation tasks and multiple face identity translation tasks. We show that our proposed approach significantly outperforms the existing domain translation methods when using no more than $10$ training samples in each image domain.

Semantics-Aligned Representation Learning for Person Re-identification

Person re-identification (reID) aims to match person images to retrieve the ones with the same identity. This is a challenging task, as the images to be matched are generally semantically misaligned due to the diversity of human poses and capture viewpoints, incompleteness of the visible bodies (due to occlusion), etc. In this paper, we propose a framework that drives the reID network to learn semantics-aligned feature representation through delicate supervision designs. Specifically, we build a Semantics Aligning Network (SAN) which consists of a base network as encoder (SA-Enc) for re-ID, and a decoder (SA-Dec) for reconstructing/regressing the densely semantics aligned full texture image. We jointly train the SAN under the supervisions of person re-identification and aligned texture generation. Moreover, at the decoder, besides the reconstruction loss, we add triplet reID constraints/losses over the feature maps as the perceptual losses. The decoder is discarded in the inference/test and thus our scheme is computationally efficient. Ablation studies demonstrate the effectiveness of our design. We achieve the state-of-the-art performances on the benchmark datasets CUHK03, Market1501, MSMT17, and the partial person reID dataset Partial REID.

Image-to-Image Translation with Multi-Path Consistency Regularization

Image translation across different domains has attracted much attention in both machine learning and computer vision communities. Taking the translation from source domain $\mathcal{D}_s$ to target domain $\mathcal{D}_t$ as an example, existing algorithms mainly rely on two kinds of loss for training: One is the discrimination loss, which is used to differentiate images generated by the models and natural images; the other is the reconstruction loss, which measures the difference between an original image and the reconstructed version through $\mathcal{D}_s\to\mathcal{D}_t\to\mathcal{D}_s$ translation. In this work, we introduce a new kind of loss, multi-path consistency loss, which evaluates the differences between direct translation $\mathcal{D}_s\to\mathcal{D}_t$ and indirect translation $\mathcal{D}_s\to\mathcal{D}_a\to\mathcal{D}_t$ with $\mathcal{D}_a$ as an auxiliary domain, to regularize training. For multi-domain translation (at least, three) which focuses on building translation models between any two domains, at each training iteration, we randomly select three domains, set them respectively as the source, auxiliary and target domains, build the multi-path consistency loss and optimize the network. For two-domain translation, we need to introduce an additional auxiliary domain and construct the multi-path consistency loss. We conduct various experiments to demonstrate the effectiveness of our proposed methods, including face-to-face translation, paint-to-photo translation, and de-raining/de-noising translation.

CaseNet: Content-Adaptive Scale Interaction Networks for Scene Parsing

Objects in an image exhibit diverse scales. Adaptive receptive fields are expected to catch suitable range of context for accurate pixel level semantic prediction for handling objects of diverse sizes. Recently, atrous convolution with different dilation rates has been used to generate features of multi-scales through several branches and these features are fused for prediction. However, there is a lack of explicit interaction among the branches to adaptively make full use of the contexts. In this paper, we propose a Content-Adaptive Scale Interaction Network (CaseNet) to exploit the multi-scale features for scene parsing. We build the CaseNet based on the classic Atrous Spatial Pyramid Pooling (ASPP) module, followed by the proposed contextual scale interaction (CSI) module, and the scale adaptation (SA) module. Specifically, first, for each spatial position, we enable context interaction among different scales through scale-aware non-local operations across the scales, \ie, CSI module, which facilitates the generation of flexible mixed receptive fields, instead of a traditional flat one. Second, the scale adaptation module (SA) explicitly and softly selects the suitable scale for each spatial position and each channel. Ablation studies demonstrate the effectiveness of the proposed modules. We achieve state-of-the-art performance on three scene parsing benchmarks Cityscapes, ADE20K and LIP.

Towards Accurate One-Stage Object Detection with AP-Loss

One-stage object detectors are trained by optimizing classification-loss and localization-loss simultaneously, with the former suffering much from extreme foreground-background class imbalance issue due to the large number of anchors. This paper alleviates this issue by proposing a novel framework to replace the classification task in one-stage detectors with a ranking task, and adopting the Average-Precision loss (AP-loss) for the ranking problem. Due to its non-differentiability and non-convexity, the AP-loss cannot be optimized directly. For this purpose, we develop a novel optimization algorithm, which seamlessly combines the error-driven update scheme in perceptron learning and backpropagation algorithm in deep networks. We verify good convergence property of the proposed algorithm theoretically and empirically. Experimental results demonstrate notable performance improvement in state-of-the-art one-stage detectors based on AP-loss over different kinds of classification-losses on various benchmarks, without changing the network architectures.

Relation-Aware Global Attention

Attention mechanism aims to increase the representation power by focusing on important features and suppressing unnecessary ones. For convolutional neural networks (CNNs), attention is typically learned with local convolutions, which ignores the global information and the hidden relation. How to efficiently exploit the long-range context to globally learn attention is underexplored. In this paper, we propose an effective Relation-Aware Global Attention (RGA) module for CNNs to fully exploit the global correlations to infer the attention. Specifically, when computing the attention at a feature position, in order to grasp information of global scope, we propose to stack the relations, i.e., its pairwise correlations/affinities with all the feature positions, and the feature itself together for learning the attention with convolutional operations. Given an intermediate feature map, we have validated the effectiveness of this design across both the spatial and channel dimensions. When applied to the task of person re-identification, our model achieves the state-of-the-art performance. Extensive ablation studies demonstrate that our RGA can significantly enhance the feature representation power. We further demonstrate the general applicability of RGA to vision tasks by applying it to the scene segmentation and image classification tasks resulting in consistent performance improvement.

Exploring Explicit Domain Supervision for Latent Space Disentanglement in Unpaired Image-to-Image Translation

Image-to-image translation tasks have been widely investigated with Generative Adversarial Networks (GANs). However, existing approaches are mostly designed in an unsupervised manner while little attention has been paid to domain information within unpaired data. In this paper, we treat domain information as explicit supervision and design an unpaired image-to-image translation framework, Domain-supervised GAN (DosGAN), which takes the first step towards the exploration of explicit domain supervision. In contrast to representing domain characteristics using different generators in CycleGAN or multiple domain codes in StarGAN, we pre-train a classification network to explicitly classify the domain of an image. After pre-training, this network is used to extract the domain-specific features of each image by using the output of its second-to-last layer. Such features, together with the domain-independent features extracted by another encoder (shared across different domains), are used to generate an image in the target domain. Extensive experiments on multiple hair color translation, multiple identity translation, multiple season translation and conditional edges-to-shoes/handbags demonstrate the effectiveness of our method. In addition, we can transfer the domain-specific feature extractor obtained on the Facescrub dataset with domain supervision information to unseen domains, such as faces in the CelebA dataset. We also succeed in achieving conditional translation with any two images in CelebA, while previous models like StarGAN cannot handle this task.