We use information-theoretic tools to derive a novel analysis of Multi-source Domain Adaptation (MDA) from the representation learning perspective. Concretely, we study joint distribution alignment for supervised MDA with few target labels and unsupervised MDA with pseudo labels, where the latter is relatively hard and less commonly studied. We further provide algorithm-dependent generalization bounds for these two settings, where the generalization is characterized by the mutual information between the parameters and the data. Then we propose a novel deep MDA algorithm, implicitly addressing the target shift through joint alignment. Finally, the mutual information bounds are extended to this algorithm providing a non-vacuous gradient-norm estimation. The proposed algorithm has comparable performance to the state-of-the-art on target-shifted MDA benchmark with improved memory efficiency.
While recent research has made significant progress in speech-driven talking face generation, the quality of the generated video still lags behind that of real recordings. One reason for this is the use of handcrafted intermediate representations like facial landmarks and 3DMM coefficients, which are designed based on human knowledge and are insufficient to precisely describe facial movements. Additionally, these methods require an external pretrained model for extracting these representations, whose performance sets an upper bound on talking face generation. To address these limitations, we propose a novel method called DAE-Talker that leverages data-driven latent representations obtained from a diffusion autoencoder (DAE). DAE contains an image encoder that encodes an image into a latent vector and a DDIM image decoder that reconstructs the image from it. We train our DAE on talking face video frames and then extract their latent representations as the training target for a Conformer-based speech2latent model. This allows DAE-Talker to synthesize full video frames and produce natural head movements that align with the content of speech, rather than relying on a predetermined head pose from a template video. We also introduce pose modelling in speech2latent for pose controllability. Additionally, we propose a novel method for generating continuous video frames with the DDIM image decoder trained on individual frames, eliminating the need for modelling the joint distribution of consecutive frames directly. Our experiments show that DAE-Talker outperforms existing popular methods in lip-sync, video fidelity, and pose naturalness. We also conduct ablation studies to analyze the effectiveness of the proposed techniques and demonstrate the pose controllability of DAE-Talker.
While generative modeling has been ubiquitous in natural language processing and computer vision, its application to image retrieval remains unexplored. In this paper, we recast image retrieval as a form of generative modeling by employing a sequence-to-sequence model, contributing to the current unified theme. Our framework, IRGen, is a unified model that enables end-to-end differentiable search, thus achieving superior performance thanks to direct optimization. While developing IRGen we tackle the key technical challenge of converting an image into quite a short sequence of semantic units in order to enable efficient and effective retrieval. Empirical experiments demonstrate that our model yields significant improvement over three commonly used benchmarks, for example, 22.9\% higher than the best baseline method in precision@10 on In-shop dataset with comparable recall@10 score.
Audio-driven talking face has attracted broad interest from academia and industry recently. However, data acquisition and labeling in audio-driven talking face are labor-intensive and costly. The lack of data resource results in poor synthesis effect. To alleviate this issue, we propose to use TTS (Text-To-Speech) for data augmentation to improve few-shot ability of the talking face system. The misalignment problem brought by the TTS audio is solved with the introduction of soft-DTW, which is first adopted in the talking face task. Moreover, features extracted by HuBERT are explored to utilize underlying information of audio, and found to be superior over other features. The proposed method achieves 17%, 14%, 38% dominance on MSE score, DTW score and user study preference repectively over the baseline model, which shows the effectiveness of improving few-shot learning for talking face system with TTS augmentation.
The explosion of short videos has dramatically reshaped the manners people socialize, yielding a new trend for daily sharing and access to the latest information. These rich video resources, on the one hand, benefited from the popularization of portable devices with cameras, but on the other, they can not be independent of the valuable editing work contributed by numerous video creators. In this paper, we investigate a novel and practical problem, namely audio beat matching (ABM), which aims to recommend the proper transition time stamps based on the background music. This technique helps to ease the labor-intensive work during video editing, saving energy for creators so that they can focus more on the creativity of video content. We formally define the ABM problem and its evaluation protocol. Meanwhile, a large-scale audio dataset, i.e., the AutoMatch with over 87k finely annotated background music, is presented to facilitate this newly opened research direction. To further lay solid foundations for the following study, we also propose a novel model termed BeatX to tackle this challenging task. Alongside, we creatively present the concept of label scope, which eliminates the data imbalance issues and assigns adaptive weights for the ground truth during the training procedure in one stop. Though plentiful short video platforms have flourished for a long time, the relevant research concerning this scenario is not sufficient, and to the best of our knowledge, AutoMatch is the first large-scale dataset to tackle the audio beat matching problem. We hope the released dataset and our competitive baseline can encourage more attention to this line of research. The dataset and codes will be made publicly available.
Automatic underground parking has attracted considerable attention as the scope of autonomous driving expands. The auto-vehicle is supposed to obtain the environmental information, track its location, and build a reliable map of the scenario. Mainstream solutions consist of well-trained neural networks and simultaneous localization and mapping (SLAM) methods, which need numerous carefully labeled images and multiple sensor estimations. However, there is a lack of underground parking scenario datasets with multiple sensors and well-labeled images that support both SLAM tasks and perception tasks, such as semantic segmentation and parking slot detection. In this paper, we present SUPS, a simulated dataset for underground automatic parking, which supports multiple tasks with multiple sensors and multiple semantic labels aligned with successive images according to timestamps. We intend to cover the defect of existing datasets with the variability of environments and the diversity and accessibility of sensors in the virtual scene. Specifically, the dataset records frames from four surrounding fisheye cameras, two forward pinhole cameras, a depth camera, and data from LiDAR, inertial measurement unit (IMU), GNSS. Pixel-level semantic labels are provided for objects, especially ground signs such as arrows, parking lines, lanes, and speed bumps. Perception, 3D reconstruction, depth estimation, and SLAM, and other relative tasks are supported by our dataset. We also evaluate the state-of-the-art SLAM algorithms and perception models on our dataset. Finally, we open source our virtual 3D scene built based on Unity Engine and release our dataset at https://github.com/jarvishou829/SUPS.
Square convolution is a default unit in convolutional neural networks as it fits well on the tensor computation for convolution operation, which usually has a fixed N x N receptive field (RF). However, what matters most to the network is the effective receptive field (ERF), which indicates the extent each pixel contributes to the output. ERF shows a Gaussian distribution and can not be modeled by simply sampling pixels with offsets. To simulate ERF, we propose a Gaussian Mask convolutional kernel (GMConv) in this work. Specifically, GMConv utilizes the Gaussian function to generate a concentric symmetry mask and put the mask over the kernel to refine the RF. Our GMConv can directly replace the standard convolutions in existing CNNs and can be easily trained end-to-end by standard backpropagation. Extensive experiments on multiple image classification benchmark datasets show that our method is comparable to, and outperforms in many cases, the standard convolution. For instance, using GMConv for AlexNet and ResNet-50, the top-1 accuracy on ImageNet classification is boosted by 0.98% and 0.85%, respectively.
DNN inference requires huge effort of system development and resource cost. This drives us to propose LUT-NN, the first trial towards empowering deep neural network (DNN) inference by table lookup, to eliminate the diverse computation kernels as well as save running cost. Based on the feature similarity of each layer, LUT-NN can learn the typical features, named centroids, of each layer from the training data, precompute them with model weights, and save the results in tables. For future input, the results of the closest centroids with the input features can be directly read from the table, as the approximation of layer output. We propose the novel centroid learning technique for DNN, which enables centroid learning through backpropagation, and adapts three levels of approximation to minimize the model loss. By this technique, LUT-NN achieves comparable accuracy (<5% difference) with original models on real complex dataset, including CIFAR, ImageNet, and GLUE. LUT-NN simplifies the computing operators to only two: closest centroid search and table lookup. We implement them for Intel and ARM CPUs. The model size is reduced by up to 3.5x for CNN models and 7x for BERT. Latency-wise, the real speedup of LUT-NN is up to 7x for BERT and 2x for ResNet, much lower than theoretical results because of the current unfriendly hardware design for table lookup. We expect firstclass table lookup support in the future to unleash the potential of LUT-NN.
Loss function learning is a new meta-learning paradigm that aims to automate the essential task of designing a loss function for a machine learning model. Existing techniques for loss function learning have shown promising results, often improving a model's training dynamics and final inference performance. However, a significant limitation of these techniques is that the loss functions are meta-learned in an offline fashion, where the meta-objective only considers the very first few steps of training, which is a significantly shorter time horizon than the one typically used for training deep neural networks. This causes significant bias towards loss functions that perform well at the very start of training but perform poorly at the end of training. To address this issue we propose a new loss function learning technique for adaptively updating the loss function online after each update to the base model parameters. The experimental results show that our proposed method consistently outperforms the cross-entropy loss and offline loss function learning techniques on a diverse range of neural network architectures and datasets.