Jianchao Yang

Xin Dong, Fuwei Zhao, Zhenyu Xie et al.

While significant progress has been made in garment transfer, one of the most applicable directions of human-centric image generation, existing works overlook the in-the-wild imagery, presenting severe garment-person misalignment as well as noticeable degradation in fine texture details. This paper, therefore, attends to virtual try-on in real-world scenes and brings essential improvements in authenticity and naturalness especially for loose garment (e.g., skirts, formal dresses), challenging poses (e.g., cross arms, bent legs), and cluttered backgrounds. Specifically, we find that the pixel flow excels at handling loose garments whereas the vertex flow is preferred for hard poses, and by combining their advantages we propose a novel generative network called wFlow that can effectively push up garment transfer to in-the-wild context. Moreover, former approaches require paired images for training. Instead, we cut down the laboriousness by working on a newly constructed large-scale video dataset named Dance50k with self-supervised cross-frame training and an online cycle optimization. The proposed Dance50k can boost real-world virtual dressing by covering a wide variety of garments under dancing poses. Extensive experiments demonstrate the superiority of our wFlow in generating realistic garment transfer results for in-the-wild images without resorting to expensive paired datasets.

Yuxi Ren, Jie Wu, Xuefeng Xiao et al.

Generative Adversarial Networks (GANs) have witnessed prevailing success in yielding outstanding images, however, they are burdensome to deploy on resource-constrained devices due to ponderous computational costs and hulking memory usage. Although recent efforts on compressing GANs have acquired remarkable results, they still exist potential model redundancies and can be further compressed. To solve this issue, we propose a novel online multi-granularity distillation (OMGD) scheme to obtain lightweight GANs, which contributes to generating high-fidelity images with low computational demands. We offer the first attempt to popularize single-stage online distillation for GAN-oriented compression, where the progressively promoted teacher generator helps to refine the discriminator-free based student generator. Complementary teacher generators and network layers provide comprehensive and multi-granularity concepts to enhance visual fidelity from diverse dimensions. Experimental results on four benchmark datasets demonstrate that OMGD successes to compress 40x MACs and 82.5X parameters on Pix2Pix and CycleGAN, without loss of image quality. It reveals that OMGD provides a feasible solution for the deployment of real-time image translation on resource-constrained devices. Our code and models are made public at: https://github.com/bytedance/OMGD.

Jieru Mei, Yingwei Li, Xiaochen Lian et al.

Search space design is very critical to neural architecture search (NAS) algorithms. We propose a fine-grained search space comprised of atomic blocks, a minimal search unit that is much smaller than the ones used in recent NAS algorithms. This search space allows a mix of operations by composing different types of atomic blocks, while the search space in previous methods only allows homogeneous operations. Based on this search space, we propose a resource-aware architecture search framework which automatically assigns the computational resources (e.g., output channel numbers) for each operation by jointly considering the performance and the computational cost. In addition, to accelerate the search process, we propose a dynamic network shrinkage technique which prunes the atomic blocks with negligible influence on outputs on the fly. Instead of a search-and-retrain two-stage paradigm, our method simultaneously searches and trains the target architecture. Our method achieves state-of-the-art performance under several FLOPs configurations on ImageNet with a small searching cost. We open our entire codebase at: https://github.com/meijieru/AtomNAS.

Yifan Jiang, Xinyu Gong, Ding Liu et al.

Deep learning-based methods have achieved remarkable success in image restoration and enhancement, but are they still competitive when there is a lack of paired training data? As one such example, this paper explores the low-light image enhancement problem, where in practice it is extremely challenging to simultaneously take a low-light and a normal-light photo of the same visual scene. We propose a highly effective unsupervised generative adversarial network, dubbed EnlightenGAN, that can be trained without low/normal-light image pairs, yet proves to generalize very well on various real-world test images. Instead of supervising the learning using ground truth data, we propose to regularize the unpaired training using the information extracted from the input itself, and benchmark a series of innovations for the low-light image enhancement problem, including a global-local discriminator structure, a self-regularized perceptual loss fusion, and attention mechanism. Through extensive experiments, our proposed approach outperforms recent methods under a variety of metrics in terms of visual quality and subjective user study. Thanks to the great flexibility brought by unpaired training, EnlightenGAN is demonstrated to be easily adaptable to enhancing real-world images from various domains. The code is available at \url{https://github.com/yueruchen/EnlightenGAN}

Jiahui Yu, Linjie Yang, Ning Xu et al.

We present a simple and general method to train a single neural network executable at different widths (number of channels in a layer), permitting instant and adaptive accuracy-efficiency trade-offs at runtime. Instead of training individual networks with different width configurations, we train a shared network with switchable batch normalization. At runtime, the network can adjust its width on the fly according to on-device benchmarks and resource constraints, rather than downloading and offloading different models. Our trained networks, named slimmable neural networks, achieve similar (and in many cases better) ImageNet classification accuracy than individually trained models of MobileNet v1, MobileNet v2, ShuffleNet and ResNet-50 at different widths respectively. We also demonstrate better performance of slimmable models compared with individual ones across a wide range of applications including COCO bounding-box object detection, instance segmentation and person keypoint detection without tuning hyper-parameters. Lastly we visualize and discuss the learned features of slimmable networks. Code and models are available at: https://github.com/JiahuiYu/slimmable_networks

Jiahui Yu, Yuchen Fan, Jianchao Yang et al.

In this report we demonstrate that with same parameters and computational budgets, models with wider features before ReLU activation have significantly better performance for single image super-resolution (SISR). The resulted SR residual network has a slim identity mapping pathway with wider (\(2\times\) to \(4\times\)) channels before activation in each residual block. To further widen activation (\(6\times\) to \(9\times\)) without computational overhead, we introduce linear low-rank convolution into SR networks and achieve even better accuracy-efficiency tradeoffs. In addition, compared with batch normalization or no normalization, we find training with weight normalization leads to better accuracy for deep super-resolution networks. Our proposed SR network \textit{WDSR} achieves better results on large-scale DIV2K image super-resolution benchmark in terms of PSNR with same or lower computational complexity. Based on WDSR, our method also won 1st places in NTIRE 2018 Challenge on Single Image Super-Resolution in all three realistic tracks. Experiments and ablation studies support the importance of wide activation for image super-resolution. Code is released at: https://github.com/JiahuiYu/wdsr_ntire2018

Yu Gao, Lixue Gong, Qiushan Guo et al.

We present Seedream 3.0, a high-performance Chinese-English bilingual image generation foundation model. We develop several technical improvements to address existing challenges in Seedream 2.0, including alignment with complicated prompts, fine-grained typography generation, suboptimal visual aesthetics and fidelity, and limited image resolutions. Specifically, the advancements of Seedream 3.0 stem from improvements across the entire pipeline, from data construction to model deployment. At the data stratum, we double the dataset using a defect-aware training paradigm and a dual-axis collaborative data-sampling framework. Furthermore, we adopt several effective techniques such as mixed-resolution training, cross-modality RoPE, representation alignment loss, and resolution-aware timestep sampling in the pre-training phase. During the post-training stage, we utilize diversified aesthetic captions in SFT, and a VLM-based reward model with scaling, thereby achieving outputs that well align with human preferences. Furthermore, Seedream 3.0 pioneers a novel acceleration paradigm. By employing consistent noise expectation and importance-aware timestep sampling, we achieve a 4 to 8 times speedup while maintaining image quality. Seedream 3.0 demonstrates significant improvements over Seedream 2.0: it enhances overall capabilities, in particular for text-rendering in complicated Chinese characters which is important to professional typography generation. In addition, it provides native high-resolution output (up to 2K), allowing it to generate images with high visual quality.

Team Seawead, Ceyuan Yang, Zhijie Lin et al.

This technical report presents a cost-efficient strategy for training a video generation foundation model. We present a mid-sized research model with approximately 7 billion parameters (7B) called Seaweed-7B trained from scratch using 665,000 H100 GPU hours. Despite being trained with moderate computational resources, Seaweed-7B demonstrates highly competitive performance compared to contemporary video generation models of much larger size. Design choices are especially crucial in a resource-constrained setting. This technical report highlights the key design decisions that enhance the performance of the medium-sized diffusion model. Empirically, we make two observations: (1) Seaweed-7B achieves performance comparable to, or even surpasses, larger models trained on substantially greater GPU resources, and (2) our model, which exhibits strong generalization ability, can be effectively adapted across a wide range of downstream applications either by lightweight fine-tuning or continue training. See the project page at https://seaweed.video/

Yu Gao, Haoyuan Guo, Tuyen Hoang et al.

Notable breakthroughs in diffusion modeling have propelled rapid improvements in video generation, yet current foundational model still face critical challenges in simultaneously balancing prompt following, motion plausibility, and visual quality. In this report, we introduce Seedance 1.0, a high-performance and inference-efficient video foundation generation model that integrates several core technical improvements: (i) multi-source data curation augmented with precision and meaningful video captioning, enabling comprehensive learning across diverse scenarios; (ii) an efficient architecture design with proposed training paradigm, which allows for natively supporting multi-shot generation and jointly learning of both text-to-video and image-to-video tasks. (iii) carefully-optimized post-training approaches leveraging fine-grained supervised fine-tuning, and video-specific RLHF with multi-dimensional reward mechanisms for comprehensive performance improvements; (iv) excellent model acceleration achieving ~10x inference speedup through multi-stage distillation strategies and system-level optimizations. Seedance 1.0 can generate a 5-second video at 1080p resolution only with 41.4 seconds (NVIDIA-L20). Compared to state-of-the-art video generation models, Seedance 1.0 stands out with high-quality and fast video generation having superior spatiotemporal fluidity with structural stability, precise instruction adherence in complex multi-subject contexts, native multi-shot narrative coherence with consistent subject representation.

Lixue Gong, Xiaoxia Hou, Fanshi Li et al.

Rapid advancement of diffusion models has catalyzed remarkable progress in the field of image generation. However, prevalent models such as Flux, SD3.5 and Midjourney, still grapple with issues like model bias, limited text rendering capabilities, and insufficient understanding of Chinese cultural nuances. To address these limitations, we present Seedream 2.0, a native Chinese-English bilingual image generation foundation model that excels across diverse dimensions, which adeptly manages text prompt in both Chinese and English, supporting bilingual image generation and text rendering. We develop a powerful data system that facilitates knowledge integration, and a caption system that balances the accuracy and richness for image description. Particularly, Seedream is integrated with a self-developed bilingual large language model as a text encoder, allowing it to learn native knowledge directly from massive data. This enable it to generate high-fidelity images with accurate cultural nuances and aesthetic expressions described in either Chinese or English. Beside, Glyph-Aligned ByT5 is applied for flexible character-level text rendering, while a Scaled ROPE generalizes well to untrained resolutions. Multi-phase post-training optimizations, including SFT and RLHF iterations, further improve the overall capability. Through extensive experimentation, we demonstrate that Seedream 2.0 achieves state-of-the-art performance across multiple aspects, including prompt-following, aesthetics, text rendering, and structural correctness. Furthermore, Seedream 2.0 has been optimized through multiple RLHF iterations to closely align its output with human preferences, as revealed by its outstanding ELO score. In addition, it can be readily adapted to an instruction-based image editing model, such as SeedEdit, with strong editing capability that balances instruction-following and image consistency.

Team Seedream, Yunpeng Chen, Yu Gao et al.

We introduce Seedream 4.0, an efficient and high-performance multimodal image generation system that unifies text-to-image (T2I) synthesis, image editing, and multi-image composition within a single framework. We develop a highly efficient diffusion transformer with a powerful VAE which also can reduce the number of image tokens considerably. This allows for efficient training of our model, and enables it to fast generate native high-resolution images (e.g., 1K-4K). Seedream 4.0 is pretrained on billions of text-image pairs spanning diverse taxonomies and knowledge-centric concepts. Comprehensive data collection across hundreds of vertical scenarios, coupled with optimized strategies, ensures stable and large-scale training, with strong generalization. By incorporating a carefully fine-tuned VLM model, we perform multi-modal post-training for training both T2I and image editing tasks jointly. For inference acceleration, we integrate adversarial distillation, distribution matching, and quantization, as well as speculative decoding. It achieves an inference time of up to 1.8 seconds for generating a 2K image (without a LLM/VLM as PE model). Comprehensive evaluations reveal that Seedream 4.0 can achieve state-of-the-art results on both T2I and multimodal image editing. In particular, it demonstrates exceptional multimodal capabilities in complex tasks, including precise image editing and in-context reasoning, and also allows for multi-image reference, and can generate multiple output images. This extends traditional T2I systems into an more interactive and multidimensional creative tool, pushing the boundary of generative AI for both creativity and professional applications. Seedream 4.0 is now accessible on https://www.volcengine.com/experience/ark?launch=seedream.

Peng Wang, Yichun Shi, Xiaochen Lian et al.

We introduce SeedEdit 3.0, in companion with our T2I model Seedream 3.0, which significantly improves over our previous SeedEdit versions in both aspects of edit instruction following and image content (e.g., ID/IP) preservation on real image inputs. Additional to model upgrading with T2I, in this report, we present several key improvements. First, we develop an enhanced data curation pipeline with a meta-info paradigm and meta-info embedding strategy that help mix images from multiple data sources. This allows us to scale editing data effectively, and meta information is helpfult to connect VLM with diffusion model more closely. Second, we introduce a joint learning pipeline for computing a diffusion loss and reward losses. Finally, we evaluate SeedEdit 3.0 on our testing benchmarks, for real/synthetic image editing, where it achieves a best trade-off between multiple aspects, yielding a high usability rate of 56.1%, compared to SeedEdit 1.6 (38.4%), GPT4o (37.1%) and Gemini 2.0 (30.3%).

Yuhan Kang, Yang Shi, Mei We et al.

Post-training pruning, as one of the key techniques for compressing large language models, plays a vital role in lightweight model deployment and model sparsity. However, current mainstream pruning methods dependent on the Hessian matrix face significant limitations in both pruning speed and practical effectiveness due to the computationally intensive nature of second-order derivative calculations. This paper presents SwiftPrune, a novel Hessian-free weight pruning method that achieves hardware-efficient model compression through two key innovations: 1) SwiftPrune eliminates the need for computationally intensive Hessian matrix calculations by introducing a contribution-based weight metric, which evaluates the importance of weights without relying on second-order derivatives. 2) we employ the Exponentially Weighted Moving Average (EWMA) technique to bypass weight sorting, enabling the selection of weights that contribute most to LLM accuracy and further reducing time complexity. Our approach is extended to support structured sparsity pruning, facilitating efficient execution on modern hardware accelerators. We validate the SwiftPrune on three LLMs (namely LLaMA2, LLaMA3, and Pythia), demonstrating that it significantly enhances compression performance. The experimental findings reveal that SwiftPrune completes the pruning process within seconds, achieving an average speedup of 12.29x (up to 56.02x) over existing SOTA approaches.

Chaosheng Dong, Xiaojie Jin, Weihao Gao et al.

Deep learning models in large-scale machine learning systems are often continuously trained with enormous data from production environments. The sheer volume of streaming training data poses a significant challenge to real-time training subsystems and ad-hoc sampling is the standard practice. Our key insight is that these deployed ML systems continuously perform forward passes on data instances during inference, but ad-hoc sampling does not take advantage of this substantial computational effort. Therefore, we propose to record a constant amount of information per instance from these forward passes. The extra information measurably improves the selection of which data instances should participate in forward and backward passes. A novel optimization framework is proposed to analyze this problem and we provide an efficient approximation algorithm under the framework of Mini-batch gradient descent as a practical solution. We also demonstrate the effectiveness of our framework and algorithm on several large-scale classification and regression tasks, when compared with competitive baselines widely used in industry.

Jian Ren, Menglei Chai, Sergey Tulyakov et al.

In this paper, we tackle the problem of human motion transfer, where we synthesize novel motion video for a target person that imitates the movement from a reference video. It is a video-to-video translation task in which the estimated poses are used to bridge two domains. Despite substantial progress on the topic, there exist several problems with the previous methods. First, there is a domain gap between training and testing pose sequences--the model is tested on poses it has not seen during training, such as difficult dancing moves. Furthermore, pose detection errors are inevitable, making the job of the generator harder. Finally, generating realistic pixels from sparse poses is challenging in a single step. To address these challenges, we introduce a novel pose-to-video translation framework for generating high-quality videos that are temporally coherent even for in-the-wild pose sequences unseen during training. We propose a pose augmentation method to minimize the training-test gap, a unified paired and unpaired learning strategy to improve the robustness to detection errors, and two-stage network architecture to achieve superior texture quality. To further boost research on the topic, we build two human motion datasets. Finally, we show the superiority of our approach over the state-of-the-art studies through extensive experiments and evaluations on different datasets.

Daquan Zhou, Xiaojie Jin, Qibin Hou et al.

The recent WSNet [1] is a new model compression method through sampling filterweights from a compact set and has demonstrated to be effective for 1D convolutionneural networks (CNNs). However, the weights sampling strategy of WSNet ishandcrafted and fixed which may severely limit the expression ability of the resultedCNNs and weaken its compression ability. In this work, we present a novel auto-sampling method that is applicable to both 1D and 2D CNNs with significantperformance improvement over WSNet. Specifically, our proposed auto-samplingmethod learns the sampling rules end-to-end instead of being independent of thenetwork architecture design. With such differentiable weight sampling rule learning,the sampling stride and channel selection from the compact set are optimized toachieve better trade-off between model compression rate and performance. Wedemonstrate that at the same compression ratio, our method outperforms WSNetby6.5% on 1D convolution. Moreover, on ImageNet, our method outperformsMobileNetV2 full model by1.47%in classification accuracy with25%FLOPsreduction. With the same backbone architecture as baseline models, our methodeven outperforms some neural architecture search (NAS) based methods such asAMC [2] and MNasNet [3].

Ning Xu, Linjie Yang, Yuchen Fan et al.

Learning long-term spatial-temporal features are critical for many video analysis tasks. However, existing video segmentation methods predominantly rely on static image segmentation techniques, and methods capturing temporal dependency for segmentation have to depend on pretrained optical flow models, leading to suboptimal solutions for the problem. End-to-end sequential learning to explore spatialtemporal features for video segmentation is largely limited by the scale of available video segmentation datasets, i.e., even the largest video segmentation dataset only contains 90 short video clips. To solve this problem, we build a new large-scale video object segmentation dataset called YouTube Video Object Segmentation dataset (YouTube-VOS). Our dataset contains 4,453 YouTube video clips and 94 object categories. This is by far the largest video object segmentation dataset to our knowledge and has been released at http://youtube-vos.org. We further evaluate several existing state-of-the-art video object segmentation algorithms on this dataset which aims to establish baselines for the development of new algorithms in the future.

Ning Xu, Linjie Yang, Yuchen Fan et al.

Learning long-term spatial-temporal features are critical for many video analysis tasks. However, existing video segmentation methods predominantly rely on static image segmentation techniques, and methods capturing temporal dependency for segmentation have to depend on pretrained optical flow models, leading to suboptimal solutions for the problem. End-to-end sequential learning to explore spatial-temporal features for video segmentation is largely limited by the scale of available video segmentation datasets, i.e., even the largest video segmentation dataset only contains 90 short video clips. To solve this problem, we build a new large-scale video object segmentation dataset called YouTube Video Object Segmentation dataset (YouTube-VOS). Our dataset contains 3,252 YouTube video clips and 78 categories including common objects and human activities. This is by far the largest video object segmentation dataset to our knowledge and we have released it at https://youtube-vos.org. Based on this dataset, we propose a novel sequence-to-sequence network to fully exploit long-term spatial-temporal information in videos for segmentation. We demonstrate that our method is able to achieve the best results on our YouTube-VOS test set and comparable results on DAVIS 2016 compared to the current state-of-the-art methods. Experiments show that the large scale dataset is indeed a key factor to the success of our model.

Jian Ren, Zhe Li, Jianchao Yang et al.

Designing the structure of neural networks is considered one of the most challenging tasks in deep learning, especially when there is few prior knowledge about the task domain. In this paper, we propose an Ecologically-Inspired GENetic (EIGEN) approach that uses the concept of succession, extinction, mimicry, and gene duplication to search neural network structure from scratch with poorly initialized simple network and few constraints forced during the evolution, as we assume no prior knowledge about the task domain. Specifically, we first use primary succession to rapidly evolve a population of poorly initialized neural network structures into a more diverse population, followed by a secondary succession stage for fine-grained searching based on the networks from the primary succession. Extinction is applied in both stages to reduce computational cost. Mimicry is employed during the entire evolution process to help the inferior networks imitate the behavior of a superior network and gene duplication is utilized to duplicate the learned blocks of novel structures, both of which help to find better network structures. Experimental results show that our proposed approach can achieve similar or better performance compared to the existing genetic approaches with dramatically reduced computation cost. For example, the network discovered by our approach on CIFAR-100 dataset achieves 78.1% test accuracy under 120 GPU hours, compared to 77.0% test accuracy in more than 65, 536 GPU hours in [35].

Jian Ren, Jianchao Yang, Ning Xu et al.

In this paper, we propose Factorized Adversarial Networks (FAN) to solve unsupervised domain adaptation problems for image classification tasks. Our networks map the data distribution into a latent feature space, which is factorized into a domain-specific subspace that contains domain-specific characteristics and a task-specific subspace that retains category information, for both source and target domains, respectively. Unsupervised domain adaptation is achieved by adversarial training to minimize the discrepancy between the distributions of two task-specific subspaces from source and target domains. We demonstrate that the proposed approach outperforms state-of-the-art methods on multiple benchmark datasets used in the literature for unsupervised domain adaptation. Furthermore, we collect two real-world tagging datasets that are much larger than existing benchmark datasets, and get significant improvement upon baselines, proving the practical value of our approach.

Linjie Yang, Yanran Wang, Xuehan Xiong et al.

Video object segmentation targets at segmenting a specific object throughout a video sequence, given only an annotated first frame. Recent deep learning based approaches find it effective by fine-tuning a general-purpose segmentation model on the annotated frame using hundreds of iterations of gradient descent. Despite the high accuracy these methods achieve, the fine-tuning process is inefficient and fail to meet the requirements of real world applications. We propose a novel approach that uses a single forward pass to adapt the segmentation model to the appearance of a specific object. Specifically, a second meta neural network named modulator is learned to manipulate the intermediate layers of the segmentation network given limited visual and spatial information of the target object. The experiments show that our approach is 70times faster than fine-tuning approaches while achieving similar accuracy.

Zhengyuan Yang, Yuncheng Li, Jianchao Yang et al.

Action recognition with 3D skeleton sequences is becoming popular due to its speed and robustness. The recently proposed Convolutional Neural Networks (CNN) based methods have shown good performance in learning spatio-temporal representations for skeleton sequences. Despite the good recognition accuracy achieved by previous CNN based methods, there exist two problems that potentially limit the performance. First, previous skeleton representations are generated by chaining joints with a fixed order. The corresponding semantic meaning is unclear and the structural information among the joints is lost. Second, previous models do not have an ability to focus on informative joints. The attention mechanism is important for skeleton based action recognition because there exist spatio-temporal key stages while the joint predictions can be inaccurate. To solve these two problems, we propose a novel CNN based method for skeleton based action recognition. We first redesign the skeleton representations with a depth-first tree traversal order, which enhances the semantic meaning of skeleton images and better preserves the associated structural information. We then propose the idea of a two-branch attention architecture that focuses on spatio-temporal key stages and filters out unreliable joint predictions. A base attention model with the simplest structure is first introduced. By improving the structures in both branches, we further propose a Global Long-sequence Attention Network (GLAN). Furthermore, in order to adjust the kernel's spatio-temporal aspect ratios and better capture long term dependencies, we propose a Sub-Sequence Attention Network (SSAN) that takes sub-image sequences as inputs. Our experiment results on NTU RGB+D and SBU Kinetic Interaction outperforms the state-of-the-art. The model is further validated on noisy estimated poses from UCF101 and Kinetics.

Yingzhen Yang, Jianchao Yang, Ning Xu et al.

We present a novel and compact architecture for deep Convolutional Neural Networks (CNNs) in this paper, termed $3$D-FilterMap Convolutional Neural Networks ($3$D-FM-CNNs). The convolution layer of $3$D-FM-CNN learns a compact representation of the filters, named $3$D-FilterMap, instead of a set of independent filters in the conventional convolution layer. The filters are extracted from the $3$D-FilterMap as overlapping $3$D submatrics with weight sharing among nearby filters, and these filters are convolved with the input to generate the output of the convolution layer for $3$D-FM-CNN. Due to the weight sharing scheme, the parameter size of the $3$D-FilterMap is much smaller than that of the filters to be learned in the conventional convolution layer when $3$D-FilterMap generates the same number of filters. Our work is fundamentally different from the network compression literature that reduces the size of a learned large network in the sense that a small network is directly learned from scratch. Experimental results demonstrate that $3$D-FM-CNN enjoys a small parameter space by learning compact $3$D-FilterMaps, while achieving performance compared to that of the baseline CNNs which learn the same number of filters as that generated by the corresponding $3$D-FilterMap.

Xiaojie Jin, Yingzhen Yang, Ning Xu et al.

We present a new approach and a novel architecture, termed WSNet, for learning compact and efficient deep neural networks. Existing approaches conventionally learn full model parameters independently and then compress them via ad hoc processing such as model pruning or filter factorization. Alternatively, WSNet proposes learning model parameters by sampling from a compact set of learnable parameters, which naturally enforces {parameter sharing} throughout the learning process. We demonstrate that such a novel weight sampling approach (and induced WSNet) promotes both weights and computation sharing favorably. By employing this method, we can more efficiently learn much smaller networks with competitive performance compared to baseline networks with equal numbers of convolution filters. Specifically, we consider learning compact and efficient 1D convolutional neural networks for audio classification. Extensive experiments on multiple audio classification datasets verify the effectiveness of WSNet. Combined with weight quantization, the resulted models are up to 180 times smaller and theoretically up to 16 times faster than the well-established baselines, without noticeable performance drop.

Xiaodan Liang, Yunchao Wei, Liang Lin et al.

An intuition on human segmentation is that when a human is moving in a video, the video-context (e.g., appearance and motion clues) may potentially infer reasonable mask information for the whole human body. Inspired by this, based on popular deep convolutional neural networks (CNN), we explore a very-weakly supervised learning framework for human segmentation task, where only an imperfect human detector is available along with massive weakly-labeled YouTube videos. In our solution, the video-context guided human mask inference and CNN based segmentation network learning iterate to mutually enhance each other until no further improvement gains. In the first step, each video is decomposed into supervoxels by the unsupervised video segmentation. The superpixels within the supervoxels are then classified as human or non-human by graph optimization with unary energies from the imperfect human detection results and the predicted confidence maps by the CNN trained in the previous iteration. In the second step, the video-context derived human masks are used as direct labels to train CNN. Extensive experiments on the challenging PASCAL VOC 2012 semantic segmentation benchmark demonstrate that the proposed framework has already achieved superior results than all previous weakly-supervised methods with object class or bounding box annotations. In addition, by augmenting with the annotated masks from PASCAL VOC 2012, our method reaches a new state-of-the-art performance on the human segmentation task.

Bowen Cheng, Zhangyang Wang, Zhaobin Zhang et al.

Emotion recognition from facial expressions is tremendously useful, especially when coupled with smart devices and wireless multimedia applications. However, the inadequate network bandwidth often limits the spatial resolution of the transmitted video, which will heavily degrade the recognition reliability. We develop a novel framework to achieve robust emotion recognition from low bit rate video. While video frames are downsampled at the encoder side, the decoder is embedded with a deep network model for joint super-resolution (SR) and recognition. Notably, we propose a novel max-mix training strategy, leading to a single "One-for-All" model that is remarkably robust to a vast range of downsampling factors. That makes our framework well adapted for the varied bandwidths in real transmission scenarios, without hampering scalability or efficiency. The proposed framework is evaluated on the AVEC 2016 benchmark, and demonstrates significantly improved stand-alone recognition performance, as well as rate-distortion (R-D) performance, than either directly recognizing from LR frames, or separating SR and recognition.

Yingzhen Yang, Jiashi Feng, Nebojsa Jojic et al.

We study the proximal gradient descent (PGD) method for $\ell^{0}$ sparse approximation problem as well as its accelerated optimization with randomized algorithms in this paper. We first offer theoretical analysis of PGD showing the bounded gap between the sub-optimal solution by PGD and the globally optimal solution for the $\ell^{0}$ sparse approximation problem under conditions weaker than Restricted Isometry Property widely used in compressive sensing literature. Moreover, we propose randomized algorithms to accelerate the optimization by PGD using randomized low rank matrix approximation (PGD-RMA) and randomized dimension reduction (PGD-RDR). Our randomized algorithms substantially reduces the computation cost of the original PGD for the $\ell^{0}$ sparse approximation problem, and the resultant sub-optimal solution still enjoys provable suboptimality, namely, the sub-optimal solution to the reduced problem still has bounded gap to the globally optimal solution to the original problem.

Yang Zhao, Ronggang Wang, Wei Jia et al.

Recent learning-based super-resolution (SR) methods often focus on dictionary learning or network training. In this paper, we discuss in detail a new SR method based on local patch encoding (LPE) instead of traditional dictionary learning. The proposed method consists of a learning stage and a reconstructing stage. In the learning stage, image patches are classified into different classes by means of the proposed LPE, and then a projection matrix is computed for each class by utilizing a simple constraint. In the reconstructing stage, an input LR patch can be simply reconstructed by computing its LPE code and then multiplying the corresponding projection matrix. Furthermore, we discuss the relationship between the proposed method and the anchored neighborhood regression methods; we also analyze the extendibility of the proposed method. The experimental results on several image sets demonstrate the effectiveness of the LPE-based methods.

Yuncheng Li, Jianchao Yang, Yale Song et al.

The ability of learning from noisy labels is very useful in many visual recognition tasks, as a vast amount of data with noisy labels are relatively easy to obtain. Traditionally, the label noises have been treated as statistical outliers, and approaches such as importance re-weighting and bootstrap have been proposed to alleviate the problem. According to our observation, the real-world noisy labels exhibit multi-mode characteristics as the true labels, rather than behaving like independent random outliers. In this work, we propose a unified distillation framework to use side information, including a small clean dataset and label relations in knowledge graph, to "hedge the risk" of learning from noisy labels. Furthermore, unlike the traditional approaches evaluated based on simulated label noises, we propose a suite of new benchmark datasets, in Sports, Species and Artifacts domains, to evaluate the task of learning from noisy labels in the practical setting. The empirical study demonstrates the effectiveness of our proposed method in all the domains.

Linjie Yang, Kevin Tang, Jianchao Yang et al.

Dense captioning is a newly emerging computer vision topic for understanding images with dense language descriptions. The goal is to densely detect visual concepts (e.g., objects, object parts, and interactions between them) from images, labeling each with a short descriptive phrase. We identify two key challenges of dense captioning that need to be properly addressed when tackling the problem. First, dense visual concept annotations in each image are associated with highly overlapping target regions, making accurate localization of each visual concept challenging. Second, the large amount of visual concepts makes it hard to recognize each of them by appearance alone. We propose a new model pipeline based on two novel ideas, joint inference and context fusion, to alleviate these two challenges. We design our model architecture in a methodical manner and thoroughly evaluate the variations in architecture. Our final model, compact and efficient, achieves state-of-the-art accuracy on Visual Genome for dense captioning with a relative gain of 73\% compared to the previous best algorithm. Qualitative experiments also reveal the semantic capabilities of our model in dense captioning.

Quanzeng You, Jiebo Luo, Hailin Jin et al.

Psychological research results have confirmed that people can have different emotional reactions to different visual stimuli. Several papers have been published on the problem of visual emotion analysis. In particular, attempts have been made to analyze and predict people's emotional reaction towards images. To this end, different kinds of hand-tuned features are proposed. The results reported on several carefully selected and labeled small image data sets have confirmed the promise of such features. While the recent successes of many computer vision related tasks are due to the adoption of Convolutional Neural Networks (CNNs), visual emotion analysis has not achieved the same level of success. This may be primarily due to the unavailability of confidently labeled and relatively large image data sets for visual emotion analysis. In this work, we introduce a new data set, which started from 3+ million weakly labeled images of different emotions and ended up 30 times as large as the current largest publicly available visual emotion data set. We hope that this data set encourages further research on visual emotion analysis. We also perform extensive benchmarking analyses on this large data set using the state of the art methods including CNNs.

Wenhan Yang, Jiashi Feng, Jianchao Yang et al.

In this work, we consider the image super-resolution (SR) problem. The main challenge of image SR is to recover high-frequency details of a low-resolution (LR) image that are important for human perception. To address this essentially ill-posed problem, we introduce a Deep Edge Guided REcurrent rEsidual~(DEGREE) network to progressively recover the high-frequency details. Different from most of existing methods that aim at predicting high-resolution (HR) images directly, DEGREE investigates an alternative route to recover the difference between a pair of LR and HR images by recurrent residual learning. DEGREE further augments the SR process with edge-preserving capability, namely the LR image and its edge map can jointly infer the sharp edge details of the HR image during the recurrent recovery process. To speed up its training convergence rate, by-pass connections across multiple layers of DEGREE are constructed. In addition, we offer an understanding on DEGREE from the view-point of sub-band frequency decomposition on image signal and experimentally demonstrate how DEGREE can recover different frequency bands separately. Extensive experiments on three benchmark datasets clearly demonstrate the superiority of DEGREE over well-established baselines and DEGREE also provides new state-of-the-arts on these datasets.

Yingzhen Yang, Jiashi Feng, Jianchao Yang et al.

Sparse subspace clustering methods, such as Sparse Subspace Clustering (SSC) \cite{ElhamifarV13} and $\ell^{1}$-graph \cite{YanW09,ChengYYFH10}, are effective in partitioning the data that lie in a union of subspaces. Most of those methods use $\ell^{1}$-norm or $\ell^{2}$-norm with thresholding to impose the sparsity of the constructed sparse similarity graph, and certain assumptions, e.g. independence or disjointness, on the subspaces are required to obtain the subspace-sparse representation, which is the key to their success. Such assumptions are not guaranteed to hold in practice and they limit the application of sparse subspace clustering on subspaces with general location. In this paper, we propose a new sparse subspace clustering method named $\ell^{0}$-graph. In contrast to the required assumptions on subspaces for most existing sparse subspace clustering methods, it is proved that subspace-sparse representation can be obtained by $\ell^{0}$-graph for arbitrary distinct underlying subspaces almost surely under the mild i.i.d. assumption on the data generation. We develop a proximal method to obtain the sub-optimal solution to the optimization problem of $\ell^{0}$-graph with proved guarantee of convergence. Moreover, we propose a regularized $\ell^{0}$-graph that encourages nearby data to have similar neighbors so that the similarity graph is more aligned within each cluster and the graph connectivity issue is alleviated. Extensive experimental results on various data sets demonstrate the superiority of $\ell^{0}$-graph compared to other competing clustering methods, as well as the effectiveness of regularized $\ell^{0}$-graph.

Quanzeng You, Jiebo Luo, Hailin Jin et al.

Sentiment analysis of online user generated content is important for many social media analytics tasks. Researchers have largely relied on textual sentiment analysis to develop systems to predict political elections, measure economic indicators, and so on. Recently, social media users are increasingly using images and videos to express their opinions and share their experiences. Sentiment analysis of such large scale visual content can help better extract user sentiments toward events or topics, such as those in image tweets, so that prediction of sentiment from visual content is complementary to textual sentiment analysis. Motivated by the needs in leveraging large scale yet noisy training data to solve the extremely challenging problem of image sentiment analysis, we employ Convolutional Neural Networks (CNN). We first design a suitable CNN architecture for image sentiment analysis. We obtain half a million training samples by using a baseline sentiment algorithm to label Flickr images. To make use of such noisy machine labeled data, we employ a progressive strategy to fine-tune the deep network. Furthermore, we improve the performance on Twitter images by inducing domain transfer with a small number of manually labeled Twitter images. We have conducted extensive experiments on manually labeled Twitter images. The results show that the proposed CNN can achieve better performance in image sentiment analysis than competing algorithms.

Xiaodan Liang, Yunchao Wei, Xiaohui Shen et al.

Instance-level object segmentation is an important yet under-explored task. The few existing studies are almost all based on region proposal methods to extract candidate segments and then utilize object classification to produce final results. Nonetheless, generating accurate region proposals itself is quite challenging. In this work, we propose a Proposal-Free Network (PFN ) to address the instance-level object segmentation problem, which outputs the instance numbers of different categories and the pixel-level information on 1) the coordinates of the instance bounding box each pixel belongs to, and 2) the confidences of different categories for each pixel, based on pixel-to-pixel deep convolutional neural network. All the outputs together, by using any off-the-shelf clustering method for simple post-processing, can naturally generate the ultimate instance-level object segmentation results. The whole PFN can be easily trained in an end-to-end way without the requirement of a proposal generation stage. Extensive evaluations on the challenging PASCAL VOC 2012 semantic segmentation benchmark demonstrate that the proposed PFN solution well beats the state-of-the-arts for instance-level object segmentation. In particular, the $AP^r$ over 20 classes at 0.5 IoU reaches 58.7% by PFN, significantly higher than 43.8% and 46.3% by the state-of-the-art algorithms, SDS [9] and [16], respectively.

Zhaowen Wang, Ding Liu, Jianchao Yang et al.

Deep learning techniques have been successfully applied in many areas of computer vision, including low-level image restoration problems. For image super-resolution, several models based on deep neural networks have been recently proposed and attained superior performance that overshadows all previous handcrafted models. The question then arises whether large-capacity and data-driven models have become the dominant solution to the ill-posed super-resolution problem. In this paper, we argue that domain expertise represented by the conventional sparse coding model is still valuable, and it can be combined with the key ingredients of deep learning to achieve further improved results. We show that a sparse coding model particularly designed for super-resolution can be incarnated as a neural network, and trained in a cascaded structure from end to end. The interpretation of the network based on sparse coding leads to much more efficient and effective training, as well as a reduced model size. Our model is evaluated on a wide range of images, and shows clear advantage over existing state-of-the-art methods in terms of both restoration accuracy and human subjective quality.

Zhangyang Wang, Jianchao Yang, Hailin Jin et al.

As font is one of the core design concepts, automatic font identification and similar font suggestion from an image or photo has been on the wish list of many designers. We study the Visual Font Recognition (VFR) problem, and advance the state-of-the-art remarkably by developing the DeepFont system. First of all, we build up the first available large-scale VFR dataset, named AdobeVFR, consisting of both labeled synthetic data and partially labeled real-world data. Next, to combat the domain mismatch between available training and testing data, we introduce a Convolutional Neural Network (CNN) decomposition approach, using a domain adaptation technique based on a Stacked Convolutional Auto-Encoder (SCAE) that exploits a large corpus of unlabeled real-world text images combined with synthetic data preprocessed in a specific way. Moreover, we study a novel learning-based model compression approach, in order to reduce the DeepFont model size without sacrificing its performance. The DeepFont system achieves an accuracy of higher than 80% (top-5) on our collected dataset, and also produces a good font similarity measure for font selection and suggestion. We also achieve around 6 times compression of the model without any visible loss of recognition accuracy.

Zhangyang Wang, Yingzhen Yang, Zhaowen Wang et al.

Deep learning has been successfully applied to image super resolution (SR). In this paper, we propose a deep joint super resolution (DJSR) model to exploit both external and self similarities for SR. A Stacked Denoising Convolutional Auto Encoder (SDCAE) is first pre-trained on external examples with proper data augmentations. It is then fine-tuned with multi-scale self examples from each input, where the reliability of self examples is explicitly taken into account. We also enhance the model performance by sub-model training and selection. The DJSR model is extensively evaluated and compared with state-of-the-arts, and show noticeable performance improvements both quantitatively and perceptually on a wide range of images.

Si Liu, Xiaodan Liang, Luoqi Liu et al.

Both parametric and non-parametric approaches have demonstrated encouraging performances in the human parsing task, namely segmenting a human image into several semantic regions (e.g., hat, bag, left arm, face). In this work, we aim to develop a new solution with the advantages of both methodologies, namely supervision from annotated data and the flexibility to use newly annotated (possibly uncommon) images, and present a quasi-parametric human parsing model. Under the classic K Nearest Neighbor (KNN)-based nonparametric framework, the parametric Matching Convolutional Neural Network (M-CNN) is proposed to predict the matching confidence and displacements of the best matched region in the testing image for a particular semantic region in one KNN image. Given a testing image, we first retrieve its KNN images from the annotated/manually-parsed human image corpus. Then each semantic region in each KNN image is matched with confidence to the testing image using M-CNN, and the matched regions from all KNN images are further fused, followed by a superpixel smoothing procedure to obtain the ultimate human parsing result. The M-CNN differs from the classic CNN in that the tailored cross image matching filters are introduced to characterize the matching between the testing image and the semantic region of a KNN image. The cross image matching filters are defined at different convolutional layers, each aiming to capture a particular range of displacements. Comprehensive evaluations over a large dataset with 7,700 annotated human images well demonstrate the significant performance gain from the quasi-parametric model over the state-of-the-arts, for the human parsing task.

Zhangyang Wang, Jianchao Yang, Hailin Jin et al.

We address a challenging fine-grain classification problem: recognizing a font style from an image of text. In this task, it is very easy to generate lots of rendered font examples but very hard to obtain real-world labeled images. This real-to-synthetic domain gap caused poor generalization to new real data in previous methods (Chen et al. (2014)). In this paper, we refer to Convolutional Neural Networks, and use an adaptation technique based on a Stacked Convolutional Auto-Encoder that exploits unlabeled real-world images combined with synthetic data. The proposed method achieves an accuracy of higher than 80% (top-5) on a real-world dataset.

Zhangyang Wang, Yingzhen Yang, Jianchao Yang et al.

We study the complementary behaviors of external and internal examples in image restoration, and are motivated to formulate a composite dictionary design framework. The composite dictionary consists of the global part learned from external examples, and the sample-specific part learned from internal examples. The dictionary atoms in both parts are further adaptively weighted to emphasize their model statistics. Experiments demonstrate that the joint utilization of external and internal examples leads to substantial improvements, with successful applications in image denoising and super resolution.

Xiaodan Liang, Si Liu, Xiaohui Shen et al.

In this work, the human parsing task, namely decomposing a human image into semantic fashion/body regions, is formulated as an Active Template Regression (ATR) problem, where the normalized mask of each fashion/body item is expressed as the linear combination of the learned mask templates, and then morphed to a more precise mask with the active shape parameters, including position, scale and visibility of each semantic region. The mask template coefficients and the active shape parameters together can generate the human parsing results, and are thus called the structure outputs for human parsing. The deep Convolutional Neural Network (CNN) is utilized to build the end-to-end relation between the input human image and the structure outputs for human parsing. More specifically, the structure outputs are predicted by two separate networks. The first CNN network is with max-pooling, and designed to predict the template coefficients for each label mask, while the second CNN network is without max-pooling to preserve sensitivity to label mask position and accurately predict the active shape parameters. For a new image, the structure outputs of the two networks are fused to generate the probability of each label for each pixel, and super-pixel smoothing is finally used to refine the human parsing result. Comprehensive evaluations on a large dataset well demonstrate the significant superiority of the ATR framework over other state-of-the-arts for human parsing. In particular, the F1-score reaches $64.38\%$ by our ATR framework, significantly higher than $44.76\%$ based on the state-of-the-art algorithm.

Zhangyang Wang, Yingzhen Yang, Zhaowen Wang et al.

Single image super-resolution (SR) aims to estimate a high-resolution (HR) image from a lowresolution (LR) input. Image priors are commonly learned to regularize the otherwise seriously ill-posed SR problem, either using external LR-HR pairs or internal similar patterns. We propose joint SR to adaptively combine the advantages of both external and internal SR methods. We define two loss functions using sparse coding based external examples, and epitomic matching based on internal examples, as well as a corresponding adaptive weight to automatically balance their contributions according to their reconstruction errors. Extensive SR results demonstrate the effectiveness of the proposed method over the existing state-of-the-art methods, and is also verified by our subjective evaluation study.

Chen Fang, Hailin Jin, Jianchao Yang et al.

Image feature representation plays an essential role in image recognition and related tasks. The current state-of-the-art feature learning paradigm is supervised learning from labeled data. However, this paradigm requires large-scale category labels, which limits its applicability to domains where labels are hard to obtain. In this paper, we propose a new data-driven feature learning paradigm which does not rely on category labels. Instead, we learn from user behavior data collected on social media. Concretely, we use the image relationship discovered in the latent space from the user behavior data to guide the image feature learning. We collect a large-scale image and user behavior dataset from Behance.net. The dataset consists of 1.9 million images and over 300 million view records from 1.9 million users. We validate our feature learning paradigm on this dataset and find that the learned feature significantly outperforms the state-of-the-art image features in learning better image similarities. We also show that the learned feature performs competitively on various recognition benchmarks.

Zhangyang Wang, Jianchao Yang, Hailin Jin et al.

We present a domain adaption framework to address a domain mismatch between synthetic training and real-world testing data. We demonstrate our method on a challenging fine-grain classification problem: recognizing a font style from an image of text. In this task, it is very easy to generate lots of rendered font examples but very hard to obtain real-world labeled images. This real-to-synthetic domain gap caused poor generalization to new real data in previous font recognition methods (Chen et al. (2014)). In this paper, we introduce a Convolutional Neural Network decomposition approach, leveraging a large training corpus of synthetic data to obtain effective features for classification. This is done using an adaptation technique based on a Stacked Convolutional Auto-Encoder that exploits a large collection of unlabeled real-world text images combined with synthetic data preprocessed in a specific way. The proposed DeepFont method achieves an accuracy of higher than 80% (top-5) on a new large labeled real-world dataset we collected.

Zhaowen Wang, Jianchao Yang, Zhe Lin et al.

Classifying large-scale image data into object categories is an important problem that has received increasing research attention. Given the huge amount of data, non-parametric approaches such as nearest neighbor classifiers have shown promising results, especially when they are underpinned by a learned distance or similarity measurement. Although metric learning has been well studied in the past decades, most existing algorithms are impractical to handle large-scale data sets. In this paper, we present an image similarity learning method that can scale well in both the number of images and the dimensionality of image descriptors. To this end, similarity comparison is restricted to each sample's local neighbors and a discriminative similarity measure is induced from large margin neighborhood embedding. We also exploit the ensemble of projections so that high-dimensional features can be processed in a set of lower-dimensional subspaces in parallel without much performance compromise. The similarity function is learned online using a stochastic gradient descent algorithm in which the triplet sampling strategy is customized for quick convergence of classification performance. The effectiveness of our proposed model is validated on several data sets with scales varying from tens of thousands to one million images. Recognition accuracies competitive with the state-of-the-art performance are achieved with much higher efficiency and scalability.

Thomas Paine, Hailin Jin, Jianchao Yang et al.

The ability to train large-scale neural networks has resulted in state-of-the-art performance in many areas of computer vision. These results have largely come from computational break throughs of two forms: model parallelism, e.g. GPU accelerated training, which has seen quick adoption in computer vision circles, and data parallelism, e.g. A-SGD, whose large scale has been used mostly in industry. We report early experiments with a system that makes use of both model parallelism and data parallelism, we call GPU A-SGD. We show using GPU A-SGD it is possible to speed up training of large convolutional neural networks useful for computer vision. We believe GPU A-SGD will make it possible to train larger networks on larger training sets in a reasonable amount of time.