A recent study has shown a phenomenon called neural collapse in that the within-class means of features and the classifier weight vectors converge to the vertices of a simplex equiangular tight frame at the terminal phase of training for classification. In this paper, we explore the corresponding structures of the last-layer feature centers and classifiers in semantic segmentation. Based on our empirical and theoretical analysis, we point out that semantic segmentation naturally brings contextual correlation and imbalanced distribution among classes, which breaks the equiangular and maximally separated structure of neural collapse for both feature centers and classifiers. However, such a symmetric structure is beneficial to discrimination for the minor classes. To preserve these advantages, we introduce a regularizer on feature centers to encourage the network to learn features closer to the appealing structure in imbalanced semantic segmentation. Experimental results show that our method can bring significant improvements on both 2D and 3D semantic segmentation benchmarks. Moreover, our method ranks 1st and sets a new record (+6.8% mIoU) on the ScanNet200 test leaderboard. Code will be available at https://github.com/dvlab-research/Imbalanced-Learning.

The surrogate loss of variational autoencoders (VAEs) poses various challenges to their training, inducing the imbalance between task fitting and representation inference. To avert this, the existing strategies for VAEs focus on adjusting the tradeoff by introducing hyperparameters, deriving a tighter bound under some mild assumptions, or decomposing the loss components per certain neural settings. VAEs still suffer from uncertain tradeoff learning.We propose a novel evolutionary variational autoencoder (eVAE) building on the variational information bottleneck (VIB) theory and integrative evolutionary neural learning. eVAE integrates a variational genetic algorithm into VAE with variational evolutionary operators including variational mutation, crossover, and evolution. Its inner-outer-joint training mechanism synergistically and dynamically generates and updates the uncertain tradeoff learning in the evidence lower bound (ELBO) without additional constraints. Apart from learning a lossy compression and representation of data under the VIB assumption, eVAE presents an evolutionary paradigm to tune critical factors of VAEs and deep neural networks and addresses the premature convergence and random search problem by integrating evolutionary optimization into deep learning. Experiments show that eVAE addresses the KL-vanishing problem for text generation with low reconstruction loss, generates all disentangled factors with sharp images, and improves the image generation quality,respectively. eVAE achieves better reconstruction loss, disentanglement, and generation-inference balance than its competitors.

Recently, large-scale pre-trained models have shown their advantages in many tasks. However, due to the huge computational complexity and storage requirements, it is challenging to apply the large-scale model to real scenes. A common solution is knowledge distillation which regards the large-scale model as a teacher model and helps to train a small student model to obtain a competitive performance. Cross-task Knowledge distillation expands the application scenarios of the large-scale pre-trained model. Existing knowledge distillation works focus on directly mimicking the final prediction or the intermediate layers of the teacher model, which represent the global-level characteristics and are task-specific. To alleviate the constraint of different label spaces, capturing invariant intrinsic local object characteristics (such as the shape characteristics of the leg and tail of the cattle and horse) plays a key role. Considering the complexity and variability of real scene tasks, we propose a Prototype-guided Cross-task Knowledge Distillation (ProC-KD) approach to transfer the intrinsic local-level object knowledge of a large-scale teacher network to various task scenarios. First, to better transfer the generalized knowledge in the teacher model in cross-task scenarios, we propose a prototype learning module to learn from the essential feature representation of objects in the teacher model. Secondly, for diverse downstream tasks, we propose a task-adaptive feature augmentation module to enhance the features of the student model with the learned generalization prototype features and guide the training of the student model to improve its generalization ability. The experimental results on various visual tasks demonstrate the effectiveness of our approach for large-scale model cross-task knowledge distillation scenes.

In this work, we propose an ID-preserving talking head generation framework, which advances previous methods in two aspects. First, as opposed to interpolating from sparse flow, we claim that dense landmarks are crucial to achieving accurate geometry-aware flow fields. Second, inspired by face-swapping methods, we adaptively fuse the source identity during synthesis, so that the network better preserves the key characteristics of the image portrait. Although the proposed model surpasses prior generation fidelity on established benchmarks, to further make the talking head generation qualified for real usage, personalized fine-tuning is usually needed. However, this process is rather computationally demanding that is unaffordable to standard users. To solve this, we propose a fast adaptation model using a meta-learning approach. The learned model can be adapted to a high-quality personalized model as fast as 30 seconds. Last but not the least, a spatial-temporal enhancement module is proposed to improve the fine details while ensuring temporal coherency. Extensive experiments prove the significant superiority of our approach over the state of the arts in both one-shot and personalized settings.

Accurate airway extraction from computed tomography (CT) images is a critical step for planning navigation bronchoscopy and quantitative assessment of airway-related chronic obstructive pulmonary disease (COPD). The existing methods are challenging to sufficiently segment the airway, especially the high-generation airway, with the constraint of the limited label and cannot meet the clinical use in COPD. We propose a novel two-stage 3D contextual transformer-based U-Net for airway segmentation using CT images. The method consists of two stages, performing initial and refined airway segmentation. The two-stage model shares the same subnetwork with different airway masks as input. Contextual transformer block is performed both in the encoder and decoder path of the subnetwork to finish high-quality airway segmentation effectively. In the first stage, the total airway mask and CT images are provided to the subnetwork, and the intrapulmonary airway mask and corresponding CT scans to the subnetwork in the second stage. Then the predictions of the two-stage method are merged as the final prediction. Extensive experiments were performed on in-house and multiple public datasets. Quantitative and qualitative analysis demonstrate that our proposed method extracted much more branches and lengths of the tree while accomplishing state-of-the-art airway segmentation performance. The code is available at https://github.com/zhaozsq/airway_segmentation.

Although considerable progress has been obtained in neural network quantization for efficient inference, existing methods are not scalable to heterogeneous devices as one dedicated model needs to be trained, transmitted, and stored for one specific hardware setting, incurring considerable costs in model training and maintenance. In this paper, we study a new vertical-layered representation of neural network weights for encapsulating all quantized models into a single one. With this representation, we can theoretically achieve any precision network for on-demand service while only needing to train and maintain one model. To this end, we propose a simple once quantization-aware training (QAT) scheme for obtaining high-performance vertical-layered models. Our design incorporates a cascade downsampling mechanism which allows us to obtain multiple quantized networks from one full precision source model by progressively mapping the higher precision weights to their adjacent lower precision counterparts. Then, with networks of different bit-widths from one source model, multi-objective optimization is employed to train the shared source model weights such that they can be updated simultaneously, considering the performance of all networks. By doing this, the shared weights will be optimized to balance the performance of different quantized models, thus making the weights transferable among different bit widths. Experiments show that the proposed vertical-layered representation and developed once QAT scheme are effective in embodying multiple quantized networks into a single one and allow one-time training, and it delivers comparable performance as that of quantized models tailored to any specific bit-width. Code will be available.

Given a piece of text, a video clip and a reference audio, the movie dubbing (also known as visual voice clone V2C) task aims to generate speeches that match the speaker's emotion presented in the video using the desired speaker voice as reference. V2C is more challenging than conventional text-to-speech tasks as it additionally requires the generated speech to exactly match the varying emotions and speaking speed presented in the video. Unlike previous works, we propose a novel movie dubbing architecture to tackle these problems via hierarchical prosody modelling, which bridges the visual information to corresponding speech prosody from three aspects: lip, face, and scene. Specifically, we align lip movement to the speech duration, and convey facial expression to speech energy and pitch via attention mechanism based on valence and arousal representations inspired by recent psychology findings. Moreover, we design an emotion booster to capture the atmosphere from global video scenes. All these embeddings together are used to generate mel-spectrogram and then convert to speech waves via existing vocoder. Extensive experimental results on the Chem and V2C benchmark datasets demonstrate the favorable performance of the proposed method. The source code and trained models will be released to the public.

The success of deep learning is partly attributed to the availability of massive data downloaded freely from the Internet. However, it also means that users' private data may be collected by commercial organizations without consent and used to train their models. Therefore, it's important and necessary to develop a method or tool to prevent unauthorized data exploitation. In this paper, we propose ConfounderGAN, a generative adversarial network (GAN) that can make personal image data unlearnable to protect the data privacy of its owners. Specifically, the noise produced by the generator for each image has the confounder property. It can build spurious correlations between images and labels, so that the model cannot learn the correct mapping from images to labels in this noise-added dataset. Meanwhile, the discriminator is used to ensure that the generated noise is small and imperceptible, thereby remaining the normal utility of the encrypted image for humans. The experiments are conducted in six image classification datasets, consisting of three natural object datasets and three medical datasets. The results demonstrate that our method not only outperforms state-of-the-art methods in standard settings, but can also be applied to fast encryption scenarios. Moreover, we show a series of transferability and stability experiments to further illustrate the effectiveness and superiority of our method.

Autoregressive language modeling (ALM) have been successfully used in self-supervised pre-training in Natural language processing (NLP). However, this paradigm has not achieved comparable results with other self-supervised approach in computer vision (e.g., contrastive learning, mask image modeling). In this paper, we try to find the reason why autoregressive modeling does not work well on vision tasks. To tackle this problem, we fully analyze the limitation of visual autoregressive methods and proposed a novel stochastic autoregressive image modeling (named SAIM) by the two simple designs. First, we employ stochastic permutation strategy to generate effective and robust image context which is critical for vision tasks. Second, we create a parallel encoder-decoder training process in which the encoder serves a similar role to the standard vision transformer focus on learning the whole contextual information, and meanwhile the decoder predicts the content of the current position, so that the encoder and decoder can reinforce each other. By introducing stochastic prediction and the parallel encoder-decoder, SAIM significantly improve the performance of autoregressive image modeling. Our method achieves the best accuracy (83.9%) on the vanilla ViT-Base model among methods using only ImageNet-1K data. Transfer performance in downstream tasks also show that our model achieves competitive performance.

Vision Transformers (ViTs) have achieved overwhelming success, yet they suffer from vulnerable resolution scalability, i.e., the performance drops drastically when presented with input resolutions that are unseen during training. We introduce, ResFormer, a framework that is built upon the seminal idea of multi-resolution training for improved performance on a wide spectrum of, mostly unseen, testing resolutions. In particular, ResFormer operates on replicated images of different resolutions and enforces a scale consistency loss to engage interactive information across different scales. More importantly, to alternate among varying resolutions, we propose a global-local positional embedding strategy that changes smoothly conditioned on input sizes. This allows ResFormer to cope with novel resolutions effectively. We conduct extensive experiments for image classification on ImageNet. The results provide strong quantitative evidence that ResFormer has promising scaling abilities towards a wide range resolutions. For instance, ResFormer-B-MR achieves a Top-1 accuracy of 75.86% and 81.72% when evaluated on relatively low and high resolutions respectively (i.e., 96 and 640), which are 48% and 7.49% better than DeiT-B. We also demonstrate, among other things, ResFormer is flexible and can be easily extended to semantic segmentation and video action recognition.