Neuroscience studies have revealed neural processes involving rhythm perception, suggesting that brain encodes rhythmic sounds and embeds information in neural activity. In this work, we investigate how to extract rhythmic information embedded in the brain responses and to decode the original audio waveforms from the extracted information. A spatiotemporal convolutional neural network is adopted to extract compact rhythm-related representations from the noninvasively measured magnetoencephalographic (MEG) signals evoked by listening to rhythmic sounds. These learned MEG representations are then used to condition an audio generator network for the synthesis of the original rhythmic sounds. In the experiments, we evaluated the proposed method by using the MEG signals recorded from eight participants and demonstrated that the generated rhythms are highly related to those evoking the MEG signals. Interestingly, we found that the auditory-related MEG channels reveal high importance in encoding rhythmic representations, the distribution of these representations relate to the timing of beats, and the behavior performance is consistent with the performance of neural decoding. These results suggest that the proposed method can synthesize rhythms by decoding neural representations from MEG.

Faces manifest large variations in many aspects, such as identity, expression, pose, and face styling. Therefore, it is a great challenge to disentangle and extract these characteristics from facial images, especially in an unsupervised manner. In this work, we introduce cycle-consistency in facial characteristics as free supervisory signal to learn facial representations from unlabeled facial images. The learning is realized by superimposing the facial motion cycle-consistency and identity cycle-consistency constraints. The main idea of the facial motion cycle-consistency is that, given a face with expression, we can perform de-expression to a neutral face via the removal of facial motion and further perform re-expression to reconstruct back to the original face. The main idea of the identity cycle-consistency is to exploit both de-identity into mean face by depriving the given neutral face of its identity via feature re-normalization and re-identity into neutral face by adding the personal attributes to the mean face. At training time, our model learns to disentangle two distinct facial representations to be useful for performing cycle-consistent face reconstruction. At test time, we use the linear protocol scheme for evaluating facial representations on various tasks, including facial expression recognition and head pose regression. We also can directly apply the learnt facial representations to person recognition, frontalization and image-to-image translation. Our experiments show that the results of our approach is competitive with those of existing methods, demonstrating the rich and unique information embedded in the dis-entangled representations. Code is available at https://github.com/JiaRenChang/FaceCycle.

The power of convolutional neural networks in arbitrary style transfer has been amply demonstrated; however, existing stylization methods tend to generate spatially inconsistent results with noticeable artifacts. One solution to this problem involves the application of a segmentation mask or affinity-based image matting to preserve spatial information related to image content. The main idea of this work is to model spatial relation between content image pixels and thus to maintain this relationship in stylization for reducing artifacts. The proposed network architecture is called spatial relation-augmented VGG (SRVGG), in which long-range spatial dependency is modeled by a spatial relation module. Based on this spatial information extracted from SRVGG, we design a novel relation loss which can minimize the difference of spatial dependency between content images and stylizations. We evaluate the proposed framework on both optimization-based and feedforward-based style transfer methods. The effectiveness of SRVGG in stylization is demonstrated by generating stylized images of high quality and spatial consistency without the need for segmentation masks or affinity-based image matting. The quantitative evaluation also suggests that the proposed framework achieve better performance compared with other methods.

Recent work has shown that depth estimation from a stereo pair of images can be formulated as a supervised learning task to be resolved with convolutional neural networks (CNNs). However, current architectures rely on patch-based Siamese networks, lacking the means to exploit context information for finding correspondence in ill-posed regions. To tackle this problem, we propose PSMNet, a pyramid stereo matching network consisting of two main modules: spatial pyramid pooling and 3D CNN. The spatial pyramid pooling module takes advantage of the capacity of global context information by aggregating context in different scales and locations to form a cost volume. The 3D CNN learns to regularize cost volume using stacked multiple hourglass networks in conjunction with intermediate supervision. The proposed approach was evaluated on several benchmark datasets. Our method ranked first in the KITTI 2012 and 2015 leaderboards before March 18, 2018. The codes of PSMNet are available at: https://github.com/JiaRenChang/PSMNet.

In the design of deep neural architectures, recent studies have demonstrated the benefits of grouping subnetworks into a larger network. For examples, the Inception architecture integrates multi-scale subnetworks and the residual network can be regarded that a residual unit combines a residual subnetwork with an identity shortcut. In this work, we embrace this observation and propose the Competitive Pathway Network (CoPaNet). The CoPaNet comprises a stack of competitive pathway units and each unit contains multiple parallel residual-type subnetworks followed by a max operation for feature competition. This mechanism enhances the model capability by learning a variety of features in subnetworks. The proposed strategy explicitly shows that the features propagate through pathways in various routing patterns, which is referred to as pathway encoding of category information. Moreover, the cross-block shortcut can be added to the CoPaNet to encourage feature reuse. We evaluated the proposed CoPaNet on four object recognition benchmarks: CIFAR-10, CIFAR-100, SVHN, and ImageNet. CoPaNet obtained the state-of-the-art or comparable results using similar amounts of parameters. The code of CoPaNet is available at: \urlhttps://github.com/JiaRenChang/CoPaNet.