Hsin-Ying Lee

Guocheng Qian, Jinjie Mai, Abdullah Hamdi et al.

We present Magic123, a two-stage coarse-to-fine approach for high-quality, textured 3D meshes generation from a single unposed image in the wild using both2D and 3D priors. In the first stage, we optimize a neural radiance field to produce a coarse geometry. In the second stage, we adopt a memory-efficient differentiable mesh representation to yield a high-resolution mesh with a visually appealing texture. In both stages, the 3D content is learned through reference view supervision and novel views guided by a combination of 2D and 3D diffusion priors. We introduce a single trade-off parameter between the 2D and 3D priors to control exploration (more imaginative) and exploitation (more precise) of the generated geometry. Additionally, we employ textual inversion and monocular depth regularization to encourage consistent appearances across views and to prevent degenerate solutions, respectively. Magic123 demonstrates a significant improvement over previous image-to-3D techniques, as validated through extensive experiments on synthetic benchmarks and diverse real-world images. Our code, models, and generated 3D assets are available at https://github.com/guochengqian/Magic123.

Dave Zhenyu Chen, Yawar Siddiqui, Hsin-Ying Lee et al.

We present Text2Tex, a novel method for generating high-quality textures for 3D meshes from the given text prompts. Our method incorporates inpainting into a pre-trained depth-aware image diffusion model to progressively synthesize high resolution partial textures from multiple viewpoints. To avoid accumulating inconsistent and stretched artifacts across views, we dynamically segment the rendered view into a generation mask, which represents the generation status of each visible texel. This partitioned view representation guides the depth-aware inpainting model to generate and update partial textures for the corresponding regions. Furthermore, we propose an automatic view sequence generation scheme to determine the next best view for updating the partial texture. Extensive experiments demonstrate that our method significantly outperforms the existing text-driven approaches and GAN-based methods.

Ligong Han, Jian Ren, Hsin-Ying Lee et al.

Most methods for conditional video synthesis use a single modality as the condition. This comes with major limitations. For example, it is problematic for a model conditioned on an image to generate a specific motion trajectory desired by the user since there is no means to provide motion information. Conversely, language information can describe the desired motion, while not precisely defining the content of the video. This work presents a multimodal video generation framework that benefits from text and images provided jointly or separately. We leverage the recent progress in quantized representations for videos and apply a bidirectional transformer with multiple modalities as inputs to predict a discrete video representation. To improve video quality and consistency, we propose a new video token trained with self-learning and an improved mask-prediction algorithm for sampling video tokens. We introduce text augmentation to improve the robustness of the textual representation and diversity of generated videos. Our framework can incorporate various visual modalities, such as segmentation masks, drawings, and partially occluded images. It can generate much longer sequences than the one used for training. In addition, our model can extract visual information as suggested by the text prompt, e.g., "an object in image one is moving northeast", and generate corresponding videos. We run evaluations on three public datasets and a newly collected dataset labeled with facial attributes, achieving state-of-the-art generation results on all four.

Yinghao Xu, Menglei Chai, Zifan Shi et al.

Existing 3D-aware image synthesis approaches mainly focus on generating a single canonical object and show limited capacity in composing a complex scene containing a variety of objects. This work presents DisCoScene: a 3Daware generative model for high-quality and controllable scene synthesis. The key ingredient of our method is a very abstract object-level representation (i.e., 3D bounding boxes without semantic annotation) as the scene layout prior, which is simple to obtain, general to describe various scene contents, and yet informative to disentangle objects and background. Moreover, it serves as an intuitive user control for scene editing. Based on such a prior, the proposed model spatially disentangles the whole scene into object-centric generative radiance fields by learning on only 2D images with the global-local discrimination. Our model obtains the generation fidelity and editing flexibility of individual objects while being able to efficiently compose objects and the background into a complete scene. We demonstrate state-of-the-art performance on many scene datasets, including the challenging Waymo outdoor dataset. Project page: https://snap-research.github.io/discoscene/

Chieh Hubert Lin, Hsin-Ying Lee, Willi Menapace et al.

Toward infinite-scale 3D city synthesis, we propose a novel framework, InfiniCity, which constructs and renders an unconstrainedly large and 3D-grounded environment from random noises. InfiniCity decomposes the seemingly impractical task into three feasible modules, taking advantage of both 2D and 3D data. First, an infinite-pixel image synthesis module generates arbitrary-scale 2D maps from the bird's-eye view. Next, an octree-based voxel completion module lifts the generated 2D map to 3D octrees. Finally, a voxel-based neural rendering module texturizes the voxels and renders 2D images. InfiniCity can thus synthesize arbitrary-scale and traversable 3D city environments, and allow flexible and interactive editing from users. We quantitatively and qualitatively demonstrate the efficacy of the proposed framework. Project page: https://hubert0527.github.io/infinicity/

Ivan Skorokhodov, Aliaksandr Siarohin, Yinghao Xu et al.

Existing 3D-from-2D generators are typically designed for well-curated single-category datasets, where all the objects have (approximately) the same scale, 3D location, and orientation, and the camera always points to the center of the scene. This makes them inapplicable to diverse, in-the-wild datasets of non-alignable scenes rendered from arbitrary camera poses. In this work, we develop a 3D generator with Generic Priors (3DGP): a 3D synthesis framework with more general assumptions about the training data, and show that it scales to very challenging datasets, like ImageNet. Our model is based on three new ideas. First, we incorporate an inaccurate off-the-shelf depth estimator into 3D GAN training via a special depth adaptation module to handle the imprecision. Then, we create a flexible camera model and a regularization strategy for it to learn its distribution parameters during training. Finally, we extend the recent ideas of transferring knowledge from pre-trained classifiers into GANs for patch-wise trained models by employing a simple distillation-based technique on top of the discriminator. It achieves more stable training than the existing methods and speeds up the convergence by at least 40%. We explore our model on four datasets: SDIP Dogs 256x256, SDIP Elephants 256x256, LSUN Horses 256x256, and ImageNet 256x256, and demonstrate that 3DGP outperforms the recent state-of-the-art in terms of both texture and geometry quality. Code and visualizations: https://snap-research.github.io/3dgp.

Tanzila Rahman, Hsin-Ying Lee, Jian Ren et al.

There has been a recent explosion of impressive generative models that can produce high quality images (or videos) conditioned on text descriptions. However, all such approaches rely on conditional sentences that contain unambiguous descriptions of scenes and main actors in them. Therefore employing such models for more complex task of story visualization, where naturally references and co-references exist, and one requires to reason about when to maintain consistency of actors and backgrounds across frames/scenes, and when not to, based on story progression, remains a challenge. In this work, we address the aforementioned challenges and propose a novel autoregressive diffusion-based framework with a visual memory module that implicitly captures the actor and background context across the generated frames. Sentence-conditioned soft attention over the memories enables effective reference resolution and learns to maintain scene and actor consistency when needed. To validate the effectiveness of our approach, we extend the MUGEN dataset and introduce additional characters, backgrounds and referencing in multi-sentence storylines. Our experiments for story generation on the MUGEN, the PororoSV and the FlintstonesSV dataset show that our method not only outperforms prior state-of-the-art in generating frames with high visual quality, which are consistent with the story, but also models appropriate correspondences between the characters and the background.

Rameen Abdal, Hsin-Ying Lee, Peihao Zhu et al.

Modern 3D-GANs synthesize geometry and texture by training on large-scale datasets with a consistent structure. Training such models on stylized, artistic data, with often unknown, highly variable geometry, and camera information has not yet been shown possible. Can we train a 3D GAN on such artistic data, while maintaining multi-view consistency and texture quality? To this end, we propose an adaptation framework, where the source domain is a pre-trained 3D-GAN, while the target domain is a 2D-GAN trained on artistic datasets. We then distill the knowledge from a 2D generator to the source 3D generator. To do that, we first propose an optimization-based method to align the distributions of camera parameters across domains. Second, we propose regularizations necessary to learn high-quality texture, while avoiding degenerate geometric solutions, such as flat shapes. Third, we show a deformation-based technique for modeling exaggerated geometry of artistic domains, enabling -- as a byproduct -- personalized geometric editing. Finally, we propose a novel inversion method for 3D-GANs linking the latent spaces of the source and the target domains. Our contributions -- for the first time -- allow for the generation, editing, and animation of personalized artistic 3D avatars on artistic datasets.

Ahmed Abdelreheem, Kyle Olszewski, Hsin-Ying Lee et al.

The two popular datasets ScanRefer [16] and ReferIt3D [3] connect natural language to real-world 3D data. In this paper, we curate a large-scale and complementary dataset extending both the aforementioned ones by associating all objects mentioned in a referential sentence to their underlying instances inside a 3D scene. Specifically, our Scan Entities in 3D (ScanEnts3D) dataset provides explicit correspondences between 369k objects across 84k natural referential sentences, covering 705 real-world scenes. Crucially, we show that by incorporating intuitive losses that enable learning from this novel dataset, we can significantly improve the performance of several recently introduced neural listening architectures, including improving the SoTA in both the Nr3D and ScanRefer benchmarks by 4.3% and 5.0%, respectively. Moreover, we experiment with competitive baselines and recent methods for the task of language generation and show that, as with neural listeners, 3D neural speakers can also noticeably benefit by training with ScanEnts3D, including improving the SoTA by 13.2 CIDEr points on the Nr3D benchmark. Overall, our carefully conducted experimental studies strongly support the conclusion that, by learning on ScanEnts3D, commonly used visio-linguistic 3D architectures can become more efficient and interpretable in their generalization without needing to provide these newly collected annotations at test time. The project's webpage is https://scanents3d.github.io/ .

Dave Zhenyu Chen, Haoxuan Li, Hsin-Ying Lee et al.

We propose SceneTex, a novel method for effectively generating high-quality and style-consistent textures for indoor scenes using depth-to-image diffusion priors. Unlike previous methods that either iteratively warp 2D views onto a mesh surface or distillate diffusion latent features without accurate geometric and style cues, SceneTex formulates the texture synthesis task as an optimization problem in the RGB space where style and geometry consistency are properly reflected. At its core, SceneTex proposes a multiresolution texture field to implicitly encode the mesh appearance. We optimize the target texture via a score-distillation-based objective function in respective RGB renderings. To further secure the style consistency across views, we introduce a cross-attention decoder to predict the RGB values by cross-attending to the pre-sampled reference locations in each instance. SceneTex enables various and accurate texture synthesis for 3D-FRONT scenes, demonstrating significant improvements in visual quality and prompt fidelity over the prior texture generation methods.

Zezhou Cheng, Menglei Chai, Jian Ren et al.

Creating and editing the shape and color of 3D objects require tremendous human effort and expertise. Compared to direct manipulation in 3D interfaces, 2D interactions such as sketches and scribbles are usually much more natural and intuitive for the users. In this paper, we propose a generic multi-modal generative model that couples the 2D modalities and implicit 3D representations through shared latent spaces. With the proposed model, versatile 3D generation and manipulation are enabled by simply propagating the editing from a specific 2D controlling modality through the latent spaces. For example, editing the 3D shape by drawing a sketch, re-colorizing the 3D surface via painting color scribbles on the 2D rendering, or generating 3D shapes of a certain category given one or a few reference images. Unlike prior works, our model does not require re-training or fine-tuning per editing task and is also conceptually simple, easy to implement, robust to input domain shifts, and flexible to diverse reconstruction on partial 2D inputs. We evaluate our framework on two representative 2D modalities of grayscale line sketches and rendered color images, and demonstrate that our method enables various shape manipulation and generation tasks with these 2D modalities.

Aniruddha Mahapatra, Aliaksandr Siarohin, Hsin-Ying Lee et al.

We introduce Text2Cinemagraph, a fully automated method for creating cinemagraphs from text descriptions - an especially challenging task when prompts feature imaginary elements and artistic styles, given the complexity of interpreting the semantics and motions of these images. We focus on cinemagraphs of fluid elements, such as flowing rivers, and drifting clouds, which exhibit continuous motion and repetitive textures. Existing single-image animation methods fall short on artistic inputs, and recent text-based video methods frequently introduce temporal inconsistencies, struggling to keep certain regions static. To address these challenges, we propose an idea of synthesizing image twins from a single text prompt - a pair of an artistic image and its pixel-aligned corresponding natural-looking twin. While the artistic image depicts the style and appearance detailed in our text prompt, the realistic counterpart greatly simplifies layout and motion analysis. Leveraging existing natural image and video datasets, we can accurately segment the realistic image and predict plausible motion given the semantic information. The predicted motion can then be transferred to the artistic image to create the final cinemagraph. Our method outperforms existing approaches in creating cinemagraphs for natural landscapes as well as artistic and other-worldly scenes, as validated by automated metrics and user studies. Finally, we demonstrate two extensions: animating existing paintings and controlling motion directions using text.

Shin-I Cheng, Yu-Jie Chen, Wei-Chen Chiu et al.

Generating images from hand-drawings is a crucial and fundamental task in content creation. The translation is difficult as there exist infinite possibilities and the different users usually expect different outcomes. Therefore, we propose a unified framework supporting a three-dimensional control over the image synthesis from sketches and strokes based on diffusion models. Users can not only decide the level of faithfulness to the input strokes and sketches, but also the degree of realism, as the user inputs are usually not consistent with the real images. Qualitative and quantitative experiments demonstrate that our framework achieves state-of-the-art performance while providing flexibility in generating customized images with control over shape, color, and realism. Moreover, our method unleashes applications such as editing on real images, generation with partial sketches and strokes, and multi-domain multi-modal synthesis.

Aliaksandr Siarohin, Willi Menapace, Ivan Skorokhodov et al.

We propose a novel approach for unsupervised 3D animation of non-rigid deformable objects. Our method learns the 3D structure and dynamics of objects solely from single-view RGB videos, and can decompose them into semantically meaningful parts that can be tracked and animated. Using a 3D autodecoder framework, paired with a keypoint estimator via a differentiable PnP algorithm, our model learns the underlying object geometry and parts decomposition in an entirely unsupervised manner. This allows it to perform 3D segmentation, 3D keypoint estimation, novel view synthesis, and animation. We primarily evaluate the framework on two video datasets: VoxCeleb $256^2$ and TEDXPeople $256^2$. In addition, on the Cats $256^2$ image dataset, we show it even learns compelling 3D geometry from still images. Finally, we show our model can obtain animatable 3D objects from a single or few images. Code and visual results available on our project website, see https://snap-research.github.io/unsupervised-volumetric-animation .

Cheng-Wei Lin, Tung-I Chen, Hsin-Ying Lee et al.

Point cloud registration is a crucial problem in computer vision and robotics. Existing methods either rely on matching local geometric features, which are sensitive to the pose differences, or leverage global shapes, which leads to inconsistency when facing distribution variances such as partial overlapping. Combining the advantages of both types of methods, we adopt a coarse-to-fine pipeline that concurrently handles both issues. We first reduce the pose differences between input point clouds by aligning global features; then we match the local features to further refine the inaccurate alignments resulting from distribution variances. As global feature alignment requires the features to preserve the poses of input point clouds and local feature matching expects the features to be invariant to these poses, we propose an SE(3)-equivariant feature extractor to simultaneously generate two types of features. In this feature extractor, representations that preserve the poses are first encoded by our novel SE(3)-equivariant network and then converted into pose-invariant ones by a pose-detaching module. Experiments demonstrate that our proposed method increases the recall rate by 20% compared to state-of-the-art methods when facing both pose differences and distribution variances.

Chieh Hubert Lin, Hsin-Ying Lee, Hung-Yu Tseng et al.

Recent studies show that paddings in convolutional neural networks encode absolute position information which can negatively affect the model performance for certain tasks. However, existing metrics for quantifying the strength of positional information remain unreliable and frequently lead to erroneous results. To address this issue, we propose novel metrics for measuring (and visualizing) the encoded positional information. We formally define the encoded information as PPP (Position-information Pattern from Padding) and conduct a series of experiments to study its properties as well as its formation. The proposed metrics measure the presence of positional information more reliably than the existing metrics based on PosENet and a test in F-Conv. We also demonstrate that for any extant (and proposed) padding schemes, PPP is primarily a learning artifact and is less dependent on the characteristics of the underlying padding schemes.

Hsin-Ying Lee, Hung-Yu Tseng, Hsin-Ying Lee et al.

Contents generated by recent advanced Text-to-Image (T2I) diffusion models are sometimes too imaginative for existing off-the-shelf dense predictors to estimate due to the immitigable domain gap. We introduce DMP, a pipeline utilizing pre-trained T2I models as a prior for dense prediction tasks. To address the misalignment between deterministic prediction tasks and stochastic T2I models, we reformulate the diffusion process through a sequence of interpolations, establishing a deterministic mapping between input RGB images and output prediction distributions. To preserve generalizability, we use low-rank adaptation to fine-tune pre-trained models. Extensive experiments across five tasks, including 3D property estimation, semantic segmentation, and intrinsic image decomposition, showcase the efficacy of the proposed method. Despite limited-domain training data, the approach yields faithful estimations for arbitrary images, surpassing existing state-of-the-art algorithms.

Yu-Jie Chen, Shin-I Cheng, Wei-Chen Chiu et al.

Current image-to-image translation methods formulate the task with conditional generation models, leading to learning only the recolorization or regional changes as being constrained by the rich structural information provided by the conditional contexts. In this work, we propose introducing the vector quantization technique into the image-to-image translation framework. The vector quantized content representation can facilitate not only the translation, but also the unconditional distribution shared among different domains. Meanwhile, along with the disentangled style representation, the proposed method further enables the capability of image extension with flexibility in both intra- and inter-domains. Qualitative and quantitative experiments demonstrate that our framework achieves comparable performance to the state-of-the-art image-to-image translation and image extension methods. Compared to methods for individual tasks, the proposed method, as a unified framework, unleashes applications combining image-to-image translation, unconditional generation, and image extension altogether. For example, it provides style variability for image generation and extension, and equips image-to-image translation with further extension capabilities.

Hsin-Ying Lee, Hung-Ting Su, Bing-Chen Tsai et al.

While recent large-scale video-language pre-training made great progress in video question answering, the design of spatial modeling of video-language models is less fine-grained than that of image-language models; existing practices of temporal modeling also suffer from weak and noisy alignment between modalities. To learn fine-grained visual understanding, we decouple spatial-temporal modeling and propose a hybrid pipeline, Decoupled Spatial-Temporal Encoders, integrating an image- and a video-language encoder. The former encodes spatial semantics from larger but sparsely sampled frames independently of time, while the latter models temporal dynamics at lower spatial but higher temporal resolution. To help the video-language model learn temporal relations for video QA, we propose a novel pre-training objective, Temporal Referring Modeling, which requires the model to identify temporal positions of events in video sequences. Extensive experiments demonstrate that our model outperforms previous work pre-trained on orders of magnitude larger datasets.

Tsung-Han Wu, Yi-Syuan Liou, Shao-Ji Yuan et al.

In the field of domain adaptation, a trade-off exists between the model performance and the number of target domain annotations. Active learning, maximizing model performance with few informative labeled data, comes in handy for such a scenario. In this work, we present D2ADA, a general active domain adaptation framework for semantic segmentation. To adapt the model to the target domain with minimum queried labels, we propose acquiring labels of the samples with high probability density in the target domain yet with low probability density in the source domain, complementary to the existing source domain labeled data. To further facilitate labeling efficiency, we design a dynamic scheduling policy to adjust the labeling budgets between domain exploration and model uncertainty over time. Extensive experiments show that our method outperforms existing active learning and domain adaptation baselines on two benchmarks, GTA5 -> Cityscapes and SYNTHIA -> Cityscapes. With less than 5% target domain annotations, our method reaches comparable results with that of full supervision. Our code is publicly available at https://github.com/tsunghan-wu/D2ADA.

Tsung-Han Wu, Yueh-Cheng Liu, Yu-Kai Huang et al.

Despite the success of deep learning on supervised point cloud semantic segmentation, obtaining large-scale point-by-point manual annotations is still a significant challenge. To reduce the huge annotation burden, we propose a Region-based and Diversity-aware Active Learning (ReDAL), a general framework for many deep learning approaches, aiming to automatically select only informative and diverse sub-scene regions for label acquisition. Observing that only a small portion of annotated regions are sufficient for 3D scene understanding with deep learning, we use softmax entropy, color discontinuity, and structural complexity to measure the information of sub-scene regions. A diversity-aware selection algorithm is also developed to avoid redundant annotations resulting from selecting informative but similar regions in a querying batch. Extensive experiments show that our method highly outperforms previous active learning strategies, and we achieve the performance of 90% fully supervised learning, while less than 15% and 5% annotations are required on S3DIS and SemanticKITTI datasets, respectively. Our code is publicly available at https://github.com/tsunghan-wu/ReDAL.

Qihang Zhang, Chaoyang Wang, Aliaksandr Siarohin et al.

We are witnessing significant breakthroughs in the technology for generating 3D objects from text. Existing approaches either leverage large text-to-image models to optimize a 3D representation or train 3D generators on object-centric datasets. Generating entire scenes, however, remains very challenging as a scene contains multiple 3D objects, diverse and scattered. In this work, we introduce SceneWiz3D, a novel approach to synthesize high-fidelity 3D scenes from text. We marry the locality of objects with globality of scenes by introducing a hybrid 3D representation: explicit for objects and implicit for scenes. Remarkably, an object, being represented explicitly, can be either generated from text using conventional text-to-3D approaches, or provided by users. To configure the layout of the scene and automatically place objects, we apply the Particle Swarm Optimization technique during the optimization process. Furthermore, it is difficult for certain parts of the scene (e.g., corners, occlusion) to receive multi-view supervision, leading to inferior geometry. We incorporate an RGBD panorama diffusion model to mitigate it, resulting in high-quality geometry. Extensive evaluation supports that our approach achieves superior quality over previous approaches, enabling the generation of detailed and view-consistent 3D scenes.

Guocheng Qian, Junli Cao, Aliaksandr Siarohin et al.

We introduce Amortized Text-to-Mesh (AToM), a feed-forward text-to-mesh framework optimized across multiple text prompts simultaneously. In contrast to existing text-to-3D methods that often entail time-consuming per-prompt optimization and commonly output representations other than polygonal meshes, AToM directly generates high-quality textured meshes in less than 1 second with around 10 times reduction in the training cost, and generalizes to unseen prompts. Our key idea is a novel triplane-based text-to-mesh architecture with a two-stage amortized optimization strategy that ensures stable training and enables scalability. Through extensive experiments on various prompt benchmarks, AToM significantly outperforms state-of-the-art amortized approaches with over 4 times higher accuracy (in DF415 dataset) and produces more distinguishable and higher-quality 3D outputs. AToM demonstrates strong generalizability, offering finegrained 3D assets for unseen interpolated prompts without further optimization during inference, unlike per-prompt solutions.

Chaoyang Wang, Peiye Zhuang, Aliaksandr Siarohin et al.

Dynamic novel view synthesis aims to capture the temporal evolution of visual content within videos. Existing methods struggle to distinguishing between motion and structure, particularly in scenarios where camera poses are either unknown or constrained compared to object motion. Furthermore, with information solely from reference images, it is extremely challenging to hallucinate unseen regions that are occluded or partially observed in the given videos. To address these issues, we first finetune a pretrained RGB-D diffusion model on the video frames using a customization technique. Subsequently, we distill the knowledge from the finetuned model to a 4D representations encompassing both dynamic and static Neural Radiance Fields (NeRF) components. The proposed pipeline achieves geometric consistency while preserving the scene identity. We perform thorough experiments to evaluate the efficacy of the proposed method qualitatively and quantitatively. Our results demonstrate the robustness and utility of our approach in challenging cases, further advancing dynamic novel view synthesis.

Bharath Raj Nagoor Kani, Hsin-Ying Lee, Sergey Tulyakov et al.

We propose UpFusion, a system that can perform novel view synthesis and infer 3D representations for an object given a sparse set of reference images without corresponding pose information. Current sparse-view 3D inference methods typically rely on camera poses to geometrically aggregate information from input views, but are not robust in-the-wild when such information is unavailable/inaccurate. In contrast, UpFusion sidesteps this requirement by learning to implicitly leverage the available images as context in a conditional generative model for synthesizing novel views. We incorporate two complementary forms of conditioning into diffusion models for leveraging the input views: a) via inferring query-view aligned features using a scene-level transformer, b) via intermediate attentional layers that can directly observe the input image tokens. We show that this mechanism allows generating high-fidelity novel views while improving the synthesis quality given additional (unposed) images. We evaluate our approach on the Co3Dv2 and Google Scanned Objects datasets and demonstrate the benefits of our method over pose-reliant sparse-view methods as well as single-view methods that cannot leverage additional views. Finally, we also show that our learned model can generalize beyond the training categories and even allow reconstruction from self-captured images of generic objects in-the-wild.

Tanzila Rahman, Shweta Mahajan, Hsin-Ying Lee et al.

Text-to-image (TTI) diffusion models have demonstrated impressive results in generating high-resolution images of complex and imaginative scenes. Recent approaches have further extended these methods with personalization techniques that allow them to integrate user-illustrated concepts (e.g., the user him/herself) using a few sample image illustrations. However, the ability to generate images with multiple interacting concepts, such as human subjects, as well as concepts that may be entangled in one, or across multiple, image illustrations remains illusive. In this work, we propose a concept-driven TTI personalization framework that addresses these core challenges. We build on existing works that learn custom tokens for user-illustrated concepts, allowing those to interact with existing text tokens in the TTI model. However, importantly, to disentangle and better learn the concepts in question, we jointly learn (latent) segmentation masks that disentangle these concepts in user-provided image illustrations. We do so by introducing an Expectation Maximization (EM)-like optimization procedure where we alternate between learning the custom tokens and estimating (latent) masks encompassing corresponding concepts in user-supplied images. We obtain these masks based on cross-attention, from within the U-Net parameterized latent diffusion model and subsequent DenseCRF optimization. We illustrate that such joint alternating refinement leads to the learning of better tokens for concepts and, as a by-product, latent masks. We illustrate the benefits of the proposed approach qualitatively and quantitatively with several examples and use cases that can combine three or more entangled concepts.

Min Jin Chong, Hsin-Ying Lee, David Forsyth

Recently, StyleGAN has enabled various image manipulation and editing tasks thanks to the high-quality generation and the disentangled latent space. However, additional architectures or task-specific training paradigms are usually required for different tasks. In this work, we take a deeper look at the spatial properties of StyleGAN. We show that with a pretrained StyleGAN along with some operations, without any additional architecture, we can perform comparably to the state-of-the-art methods on various tasks, including image blending, panorama generation, generation from a single image, controllable and local multimodal image to image translation, and attributes transfer. The proposed method is simple, effective, efficient, and applicable to any existing pretrained StyleGAN model.

Yan-Bo Lin, Hung-Yu Tseng, Hsin-Ying Lee et al.

Sound localization aims to find the source of the audio signal in the visual scene. However, it is labor-intensive to annotate the correlations between the signals sampled from the audio and visual modalities, thus making it difficult to supervise the learning of a machine for this task. In this work, we propose an iterative contrastive learning framework that requires no data annotations. At each iteration, the proposed method takes the 1) localization results in images predicted in the previous iteration, and 2) semantic relationships inferred from the audio signals as the pseudo-labels. We then use the pseudo-labels to learn the correlation between the visual and audio signals sampled from the same video (intra-frame sampling) as well as the association between those extracted across videos (inter-frame relation). Our iterative strategy gradually encourages the localization of the sounding objects and reduces the correlation between the non-sounding regions and the reference audio. Quantitative and qualitative experimental results demonstrate that the proposed framework performs favorably against existing unsupervised and weakly-supervised methods on the sound localization task.

Yu-Ding Lu, Hsin-Ying Lee, Hung-Yu Tseng et al.

As recent generative models can generate photo-realistic images, people seek to understand the mechanism behind the generation process. Interpretable generation process is beneficial to various image editing applications. In this work, we propose a framework to discover interpretable directions in the latent space given arbitrary pre-trained generative adversarial networks. We propose to learn the transformation from prior one-hot vectors representing different attributes to the latent space used by pre-trained models. Furthermore, we apply a centroid loss function to improve consistency and smoothness while traversing through different directions. We demonstrate the efficacy of the proposed framework on a wide range of datasets. The discovered direction vectors are shown to be visually corresponding to various distinct attributes and thus enable attribute editing.

Hung-Yu Tseng, Hsin-Ying Lee, Lu Jiang et al.

Image generation from scene description is a cornerstone technique for the controlled generation, which is beneficial to applications such as content creation and image editing. In this work, we aim to synthesize images from scene description with retrieved patches as reference. We propose a differentiable retrieval module. With the differentiable retrieval module, we can (1) make the entire pipeline end-to-end trainable, enabling the learning of better feature embedding for retrieval; (2) encourage the selection of mutually compatible patches with additional objective functions. We conduct extensive quantitative and qualitative experiments to demonstrate that the proposed method can generate realistic and diverse images, where the retrieved patches are reasonable and mutually compatible.

Yen-Chi Cheng, Hsin-Ying Lee, Min Sun et al.

Flexible user controls are desirable for content creation and image editing. A semantic map is commonly used intermediate representation for conditional image generation. Compared to the operation on raw RGB pixels, the semantic map enables simpler user modification. In this work, we specifically target at generating semantic maps given a label-set consisting of desired categories. The proposed framework, SegVAE, synthesizes semantic maps in an iterative manner using conditional variational autoencoder. Quantitative and qualitative experiments demonstrate that the proposed model can generate realistic and diverse semantic maps. We also apply an off-the-shelf image-to-image translation model to generate realistic RGB images to better understand the quality of the synthesized semantic maps. Furthermore, we showcase several real-world image-editing applications including object removal, object insertion, and object replacement.

Jia-Fong Yeh, Hsin-Ying Lee, Bing-Chen Tsai et al.

In recent years, few-shot learning problems have received a lot of attention. While methods in most previous works were trained and tested on datasets in one single domain, cross-domain few-shot learning is a brand-new branch of few-shot learning problems, where models handle datasets in different domains between training and testing phases. In this paper, to solve the problem that the model is pre-trained (meta-trained) on a single dataset while fine-tuned on datasets in four different domains, including common objects, satellite images, and medical images, we propose a novel large margin fine-tuning method (LMM-PQS), which generates pseudo query images from support images and fine-tunes the feature extraction modules with a large margin mechanism inspired by methods in face recognition. According to the experiment results, LMM-PQS surpasses the baseline models by a significant margin and demonstrates that our approach is robust and can easily adapt pre-trained models to new domains with few data.

Hung-Yu Tseng, Hsin-Ying Lee, Jia-Bin Huang et al.

Few-shot classification aims to recognize novel categories with only few labeled images in each class. Existing metric-based few-shot classification algorithms predict categories by comparing the feature embeddings of query images with those from a few labeled images (support examples) using a learned metric function. While promising performance has been demonstrated, these methods often fail to generalize to unseen domains due to large discrepancy of the feature distribution across domains. In this work, we address the problem of few-shot classification under domain shifts for metric-based methods. Our core idea is to use feature-wise transformation layers for augmenting the image features using affine transforms to simulate various feature distributions under different domains in the training stage. To capture variations of the feature distributions under different domains, we further apply a learning-to-learn approach to search for the hyper-parameters of the feature-wise transformation layers. We conduct extensive experiments and ablation studies under the domain generalization setting using five few-shot classification datasets: mini-ImageNet, CUB, Cars, Places, and Plantae. Experimental results demonstrate that the proposed feature-wise transformation layer is applicable to various metric-based models, and provides consistent improvements on the few-shot classification performance under domain shift.

Hsin-Ying Lee, Lu Jiang, Irfan Essa et al.

Graphic design is essential for visual communication with layouts being fundamental to composing attractive designs. Layout generation differs from pixel-level image synthesis and is unique in terms of the requirement of mutual relations among the desired components. We propose a method for design layout generation that can satisfy user-specified constraints. The proposed neural design network (NDN) consists of three modules. The first module predicts a graph with complete relations from a graph with user-specified relations. The second module generates a layout from the predicted graph. Finally, the third module fine-tunes the predicted layout. Quantitative and qualitative experiments demonstrate that the generated layouts are visually similar to real design layouts. We also construct real designs based on predicted layouts for a better understanding of the visual quality. Finally, we demonstrate a practical application on layout recommendation.

Yu-Ding Lu, Hsin-Ying Lee, Hung-Yu Tseng et al.

Spatial audio is an essential medium to audiences for 3D visual and auditory experience. However, the recording devices and techniques are expensive or inaccessible to the general public. In this work, we propose a self-supervised audio spatialization network that can generate spatial audio given the corresponding video and monaural audio. To enhance spatialization performance, we use an auxiliary classifier to classify ground-truth videos and those with audio where the left and right channels are swapped. We collect a large-scale video dataset with spatial audio to validate the proposed method. Experimental results demonstrate the effectiveness of the proposed model on the audio spatialization task.

Hsin-Ying Lee, Hung-Yu Tseng, Qi Mao et al.

Image-to-image translation aims to learn the mapping between two visual domains. There are two main challenges for this task: 1) lack of aligned training pairs and 2) multiple possible outputs from a single input image. In this work, we present an approach based on disentangled representation for generating diverse outputs without paired training images. To synthesize diverse outputs, we propose to embed images onto two spaces: a domain-invariant content space capturing shared information across domains and a domain-specific attribute space. Our model takes the encoded content features extracted from a given input and attribute vectors sampled from the attribute space to synthesize diverse outputs at test time. To handle unpaired training data, we introduce a cross-cycle consistency loss based on disentangled representations. Qualitative results show that our model can generate diverse and realistic images on a wide range of tasks without paired training data. For quantitative evaluations, we measure realism with user study and Fréchet inception distance, and measure diversity with the perceptual distance metric, Jensen-Shannon divergence, and number of statistically-different bins.

Qi Mao, Hsin-Ying Lee, Hung-Yu Tseng et al.

Most conditional generation tasks expect diverse outputs given a single conditional context. However, conditional generative adversarial networks (cGANs) often focus on the prior conditional information and ignore the input noise vectors, which contribute to the output variations. Recent attempts to resolve the mode collapse issue for cGANs are usually task-specific and computationally expensive. In this work, we propose a simple yet effective regularization term to address the mode collapse issue for cGANs. The proposed method explicitly maximizes the ratio of the distance between generated images with respect to the corresponding latent codes, thus encouraging the generators to explore more minor modes during training. This mode seeking regularization term is readily applicable to various conditional generation tasks without imposing training overhead or modifying the original network structures. We validate the proposed algorithm on three conditional image synthesis tasks including categorical generation, image-to-image translation, and text-to-image synthesis with different baseline models. Both qualitative and quantitative results demonstrate the effectiveness of the proposed regularization method for improving diversity without loss of quality.

Hsin-Ying Lee, Hung-Yu Tseng, Jia-Bin Huang et al.

Image-to-image translation aims to learn the mapping between two visual domains. There are two main challenges for many applications: 1) the lack of aligned training pairs and 2) multiple possible outputs from a single input image. In this work, we present an approach based on disentangled representation for producing diverse outputs without paired training images. To achieve diversity, we propose to embed images onto two spaces: a domain-invariant content space capturing shared information across domains and a domain-specific attribute space. Our model takes the encoded content features extracted from a given input and the attribute vectors sampled from the attribute space to produce diverse outputs at test time. To handle unpaired training data, we introduce a novel cross-cycle consistency loss based on disentangled representations. Qualitative results show that our model can generate diverse and realistic images on a wide range of tasks without paired training data. For quantitative comparisons, we measure realism with user study and diversity with a perceptual distance metric. We apply the proposed model to domain adaptation and show competitive performance when compared to the state-of-the-art on the MNIST-M and the LineMod datasets.

Hsin-Ying Lee, Jia-Bin Huang, Maneesh Singh et al.

We present an unsupervised representation learning approach using videos without semantic labels. We leverage the temporal coherence as a supervisory signal by formulating representation learning as a sequence sorting task. We take temporally shuffled frames (i.e., in non-chronological order) as inputs and train a convolutional neural network to sort the shuffled sequences. Similar to comparison-based sorting algorithms, we propose to extract features from all frame pairs and aggregate them to predict the correct order. As sorting shuffled image sequence requires an understanding of the statistical temporal structure of images, training with such a proxy task allows us to learn rich and generalizable visual representation. We validate the effectiveness of the learned representation using our method as pre-training on high-level recognition problems. The experimental results show that our method compares favorably against state-of-the-art methods on action recognition, image classification and object detection tasks.

Dong Li, Hsin-Ying Lee, Jia-Bin Huang et al.

Numerous embedding models have been recently explored to incorporate semantic knowledge into visual recognition. Existing methods typically focus on minimizing the distance between the corresponding images and texts in the embedding space but do not explicitly optimize the underlying structure. Our key observation is that modeling the pairwise image-image relationship improves the discrimination ability of the embedding model. In this paper, we propose the structured discriminative and difference constraints to learn visual-semantic embeddings. First, we exploit the discriminative constraints to capture the intra- and inter-class relationships of image embeddings. The discriminative constraints encourage separability for image instances of different classes. Second, we align the difference vector between a pair of image embeddings with that of the corresponding word embeddings. The difference constraints help regularize image embeddings to preserve the semantic relationships among word embeddings. Extensive evaluations demonstrate the effectiveness of the proposed structured embeddings for single-label classification, multi-label classification, and zero-shot recognition.