Disentangled generative models are typically trained with an extra regularization term, which encourages the traversal of each latent factor to make a distinct and independent change at the cost of generation quality. When traversing the latent space of generative models trained without the disentanglement term, the generated samples show semantically meaningful change, raising the question: do generative models know disentanglement? We propose an unsupervised and model-agnostic method: Disentanglement via Contrast (DisCo) in the Variation Space. DisCo consists of: (i) a Navigator providing traversal directions in the latent space, and (ii) a $\Delta$-Contrastor composed of two shared-weight Encoders, which encode image pairs along these directions to disentangled representations respectively, and a difference operator to map the encoded representations to the Variation Space. We propose two more key techniques for DisCo: entropy-based domination loss to make the encoded representations more disentangled and the strategy of flipping hard negatives to address directions with the same semantic meaning. By optimizing the Navigator to discover disentangled directions in the latent space and Encoders to extract disentangled representations from images with Contrastive Learning, DisCo achieves the state-of-the-art disentanglement given pretrained non-disentangled generative models, including GAN, VAE, and Flow. Project page at https://github.com/xrenaa/DisCo.
The key idea of the state-of-the-art VAE-based unsupervised representation disentanglement methods is to minimize the total correlation of the latent variable distributions. However, it has been proved that VAE-based unsupervised disentanglement can not be achieved without introducing other inductive bias. In this paper, we address VAE-based unsupervised disentanglement by leveraging the constraints derived from the Group Theory based definition as the non-probabilistic inductive bias. More specifically, inspired by the nth dihedral group (the permutation group for regular polygons), we propose a specific form of the definition and prove its two equivalent conditions: isomorphism and "the constancy of permutations". We further provide an implementation of isomorphism based on two Group constraints: the Abel constraint for the exchangeability and Order constraint for the cyclicity. We then convert them into a self-supervised training loss that can be incorporated into VAE-based models to bridge their gaps from the Group Theory based definition. We train 1800 models covering the most prominent VAE-based models on five datasets to verify the effectiveness of our method. Compared to the original models, the Groupidied VAEs consistently achieve better mean performance with smaller variances, and make meaningful dimensions controllable.
Vehicle Re-Identification (V-ReID) is a critical task that associates the same vehicle across images from different camera viewpoints. Many works explore attribute clues to enhance V-ReID; however, there is usually a lack of effective interaction between the attribute-related modules and final V-ReID objective. In this work, we propose a new method to efficiently explore discriminative information from vehicle attributes (e.g., color and type). We introduce AttributeNet (ANet) that jointly extracts identity-relevant features and attribute features. We enable the interaction by distilling the ReID-helpful attribute feature and adding it into the general ReID feature to increase the discrimination power. Moreover, we propose a constraint, named Amelioration Constraint (AC), which encourages the feature after adding attribute features onto the general ReID feature to be more discriminative than the original general ReID feature. We validate the effectiveness of our framework on three challenging datasets. Experimental results show that our method achieves state-of-the-art performance.
This paper presents a self-supervised learning framework, named MGF, for general-purpose speech representation learning. In the design of MGF, speech hierarchy is taken into consideration. Specifically, we propose to use generative learning approaches to capture fine-grained information at small time scales and use discriminative learning approaches to distill coarse-grained or semantic information at large time scales. For phoneme-scale learning, we borrow idea from the masked language model but tailor it for the continuous speech signal by replacing classification loss with a contrastive loss. We corroborate our design by evaluating MGF representation on various downstream tasks, including phoneme classification, speaker classification, speech recognition, and emotion classification. Experiments verify that training at different time scales needs different training targets and loss functions, which in general complement each other and lead to a better performance.
Predicting future frames of video sequences is challenging due to the complex and stochastic nature of the problem. Video prediction methods based on variational auto-encoders (VAEs) have been a great success, but they require the training data to contain multiple possible futures for an observed video sequence. This is hard to be fulfilled when videos are captured in the wild where any given observation only has a determinate future. As a result, training a vanilla VAE model with these videos inevitably causes posterior collapse. To alleviate this problem, we propose a novel VAE structure, dabbed VAE-in-VAE or VAE$^2$. The key idea is to explicitly introduce stochasticity into the VAE. We treat part of the observed video sequence as a random transition state that bridges its past and future, and maximize the likelihood of a Markov Chain over the video sequence under all possible transition states. A tractable lower bound is proposed for this intractable objective function and an end-to-end optimization algorithm is designed accordingly. VAE$^2$ can mitigate the posterior collapse problem to a large extent, as it breaks the direct dependence between future and observation and does not directly regress the determinate future provided by the training data. We carry out experiments on a large-scale dataset called Cityscapes, which contains videos collected from a number of urban cities. Results show that VAE$^2$ is capable of predicting diverse futures and is more resistant to posterior collapse than the other state-of-the-art VAE-based approaches. We believe that VAE$^2$ is also applicable to other stochastic sequence prediction problems where training data are lack of stochasticity.
For many practical computer vision applications, the learned models usually have high performance on the datasets used for training but suffer from significant performance degradation when deployed in new environments, where there are usually style differences between the training images and the testing images. An effective domain generalizable model is expected to be able to learn feature representations that are both generalizable and discriminative. In this paper, we design a novel Style Normalization and Restitution module (SNR) to simultaneously ensure both high generalization and discrimination capability of the networks. In the SNR module, particularly, we filter out the style variations (e.g, illumination, color contrast) by performing Instance Normalization (IN) to obtain style normalized features, where the discrepancy among different samples and domains is reduced. However, such a process is task-ignorant and inevitably removes some task-relevant discriminative information, which could hurt the performance. To remedy this, we propose to distill task-relevant discriminative features from the residual (i.e, the difference between the original feature and the style normalized feature) and add them back to the network to ensure high discrimination. Moreover, for better disentanglement, we enforce a dual causality loss constraint in the restitution step to encourage the better separation of task-relevant and task-irrelevant features. We validate the effectiveness of our SNR on different computer vision tasks, including classification, semantic segmentation, and object detection. Experiments demonstrate that our SNR module is capable of improving the performance of networks for domain generalization (DG) and unsupervised domain adaptation (UDA) on many tasks. Code are available at https://github.com/microsoft/SNR.
Many unsupervised domain adaptive (UDA) person re-identification (ReID) approaches combine clustering-based pseudo-label prediction with feature fine-tuning. However, because of domain gap, the pseudo-labels are not always reliable and there are noisy/incorrect labels. This would mislead the feature representation learning and deteriorate the performance. In this paper, we propose to estimate and exploit the credibility of the assigned pseudo-label of each sample to alleviate the influence of noisy labels, by suppressing the contribution of noisy samples. We build our baseline framework using the mean teacher method together with an additional contrastive loss. We have observed that a sample with a wrong pseudo-label through clustering in general has a weaker consistency between the output of the mean teacher model and the student model. Based on this finding, we propose to exploit the uncertainty (measured by consistency levels) to evaluate the reliability of the pseudo-label of a sample and incorporate the uncertainty to re-weight its contribution within various ReID losses, including the identity (ID) classification loss per sample, the triplet loss, and the contrastive loss. Our uncertainty-guided optimization brings significant improvement and achieves the state-of-the-art performance on benchmark datasets.
Semi-supervised learning aims to boost the accuracy of a model by exploring unlabeled images. The state-of-the-art methods are consistency-based which learn about unlabeled images by encouraging the model to give consistent predictions for images under different augmentations. However, when applied to pose estimation, the methods degenerate and predict every pixel in unlabeled images as background. This is because contradictory predictions are gradually pushed to the background class due to highly imbalanced class distribution. But this is not an issue in supervised learning because it has accurate labels. This inspires us to stabilize the training by obtaining reliable pseudo labels. Specifically, we learn two networks to mutually teach each other. In particular, for each image, we compose an easy-hard pair by applying different augmentations and feed them to both networks. The more reliable predictions on easy images in each network are used to teach the other network to learn about the corresponding hard images. The approach successfully avoids degeneration and achieves promising results on public datasets. The source code will be released.
Re-identification (re-ID) is currently investigated as a closed-world image retrieval task, and evaluated by retrieval based metrics. The algorithms return ranking lists to users, but cannot tell which images are the true target. In essence, current re-ID overemphasizes the importance of retrieval but underemphasizes that of verification, \textit{i.e.}, all returned images are considered as the target. On the other hand, re-ID should also include the scenario that the query identity does not appear in the gallery. To this end, we go back to the essence of re-ID, \textit{i.e.}, a combination of retrieval and verification in an open-set setting, and put forward a new metric, namely, Genuine Open-set re-ID Metric (GOM). GOM explicitly balances the effect of performing retrieval and verification into a single unified metric. It can also be decomposed into a family of sub-metrics, enabling a clear analysis of re-ID performance. We evaluate the effectiveness of GOM on the re-ID benchmarks, showing its ability to capture important aspects of re-ID performance that have not been taken into account by established metrics so far. Furthermore, we show GOM scores excellent in aligning with human visual evaluation of re-ID performance. Related codes are available at https://github.com/YuanXinCherry/Person-reID-Evaluation
Occlusion is probably the biggest challenge for human pose estimation in the wild. Typical solutions often rely on intrusive sensors such as IMUs to detect occluded joints. To make the task truly unconstrained, we present AdaFuse, an adaptive multiview fusion method, which can enhance the features in occluded views by leveraging those in visible views. The core of AdaFuse is to determine the point-point correspondence between two views which we solve effectively by exploring the sparsity of the heatmap representation. We also learn an adaptive fusion weight for each camera view to reflect its feature quality in order to reduce the chance that good features are undesirably corrupted by ``bad'' views. The fusion model is trained end-to-end with the pose estimation network, and can be directly applied to new camera configurations without additional adaptation. We extensively evaluate the approach on three public datasets including Human3.6M, Total Capture and CMU Panoptic. It outperforms the state-of-the-arts on all of them. We also create a large scale synthetic dataset Occlusion-Person, which allows us to perform numerical evaluation on the occluded joints, as it provides occlusion labels for every joint in the images. The dataset and code are released at https://github.com/zhezh/adafuse-3d-human-pose.