skip to main content


Title: FineGAN: Unsupervised Hierarchical Disentanglement for Fine-Grained Object Generation and Discovery
We propose FineGAN, a novel unsupervised GAN framework, which disentangles the background, object shape, and object appearance to hierarchically generate images of fine-grained object categories. To disentangle the factors without supervision, our key idea is to use information theory to associate each factor to a latent code, and to condition the relationships between the codes in a specific way to induce the desired hierarchy. Through extensive experiments, we show that FineGAN achieves the desired disentanglement to generate realistic and diverse images belonging to fine-grained classes of birds, dogs, and cars. Using FineGAN’s automatically learned features, we also cluster real images as a first attempt at solving the novel problem of unsupervised fine-grained object category discovery. Our code/models/demo can be found at https://github.com/kkanshul/finegan  more » « less
Award ID(s):
1748387
NSF-PAR ID:
10100527
Author(s) / Creator(s):
; ;
Date Published:
Journal Name:
IEEE Conference on Computer Vision and Pattern Recognition (CVPR)
Format(s):
Medium: X
Sponsoring Org:
National Science Foundation
More Like this
  1. We propose FineGAN, a novel unsupervised GAN framework, which disentangles the background, object shape, and object appearance to hierarchically generate images of fine-grained object categories. To disentangle the factors without any supervision, our key idea is to use information theory to associate each factor to a latent code, and to condition the relationships between the codes in a specific way to induce the desired hierarchy. Through extensive experiments, we show that FineGAN achieves the desired disentanglement to generate realistic and diverse images belonging to fine-grained classes of birds, dogs, and cars. Using FineGAN's automatically learned features, we also cluster real images as a first attempt at solving the novel problem of unsupervised fine-grained object category discovery. 
    more » « less
  2. Despite remarkable recent progress on both unconditional and conditional image synthesis, it remains a long-standing problem to learn generative models that are capable of synthesizing realistic and sharp images from reconfigurable spatial layout (i.e., bounding boxes + class labels in an image lattice) and style (i.e., structural and appearance variations encoded by latent vectors), especially at high resolution. By reconfigurable, it means that a model can preserve the intrinsic one-to-many mapping from a given layout to multiple plausible images with different styles, and is adaptive with respect to perturbations of a layout and style latent code. In this paper, we present a layout- and style-based architecture for generative adversarial networks (termed LostGANs) that can be trained end-to-end to generate images from reconfigurable layout and style. Inspired by the vanilla StyleGAN, the proposed LostGAN consists of two new components: (i) learning fine-grained mask maps in a weakly-supervised manner to bridge the gap between layouts and images, and (ii) learning object instance-specific layout-aware feature normalization (ISLA-Norm) in the generator to realize multi-object style generation. In experiments, the proposed method is tested on the COCO-Stuff dataset and the Visual Genome dataset with state-of-the-art performance obtained. The code and pretrained models are available at https://github.com/iVMCL/LostGANs 
    more » « less
  3. Vision-language (VL) pre-training has recently received considerable attention. However, most existing end-to-end pre-training approaches either only aim to tackle VL tasks such as image-text retrieval, visual question answering (VQA) and image captioning that test high-level understanding of images, or only target region-level understanding for tasks such as phrase grounding and object detection. We present FIBER (Fusion-In-the-Backbone-based transformER), a new VL model architecture that can seamlessly handle both these types of tasks. Instead of having dedicated transformer layers for fusion after the uni-modal backbones, FIBER pushes multimodal fusion deep into the model by inserting cross-attention into the image and text backbones to better capture multimodal interactions. In addition, unlike previous work that is either only pre-trained on image-text data or on fine-grained data with box-level annotations, we present a two-stage pre-training strategy that uses both these kinds of data efficiently: (i) coarse-grained pre-training based on image-text data; followed by (ii) fine-grained pre-training based on image-text-box data. We conduct comprehensive experiments on a wide range of VL tasks, ranging from VQA, image captioning, and retrieval, to phrase grounding, referring expression comprehension, and object detection. Using deep multimodal fusion coupled with the two-stage pre-training, FIBER provides consistent performance improvements over strong baselines across all tasks, often outperforming methods using magnitudes more data. Code is released at https://github.com/microsoft/FIBER. 
    more » « less
  4. null (Ed.)
    Human novel view synthesis aims to synthesize target views of a human subject given input images taken from one or more reference viewpoints. Despite significant advances in model-free novel view synthesis, existing methods present two major limitations when applied to complex shapes like humans. First, these methods mainly focus on simple and symmetric objects, e.g., cars and chairs, limiting their performances to fine-grained and asymmetric shapes. Second, existing methods cannot guarantee visual consistency across different adjacent views of the same object. To solve these problems, we present in this paper a learning framework for the novel view synthesis of human subjects, which explicitly enforces consistency across different generated views of the subject. Specifically, we introduce a novel multi-view supervision and an explicit rotational loss during the learning process, enabling the model to preserve detailed body parts and to achieve consistency between adjacent synthesized views. To show the superior performance of our approach, we present qualitative and quantitative results on the Multi-View Human Action (MVHA) dataset we collected (consisting of 3D human models animated with different Mocap sequences and captured from 54 different viewpoints), the Pose-Varying Human Model (PVHM) dataset, and ShapeNet. The qualitative and quantitative results demonstrate that our approach outperforms the state-of-the-art baselines in both per-view synthesis quality, and in preserving rotational consistency and complex shapes (e.g. fine-grained details, challenging poses) across multiple adjacent views in a variety of scenarios, for both humans and rigid objects. 
    more » « less
  5. As augmented and virtual reality (AR/VR) technology matures, a method is desired to represent real-world persons visually and aurally in a virtual scene with high fidelity to craft an immersive and realistic user experience. Current technologies leverage camera and depth sensors to render visual representations of subjects through avatars, and microphone arrays are employed to localize and separate high-quality subject audio through beamforming. However, challenges remain in both realms. In the visual domain, avatars can only map key features (e.g., pose, expression) to a predetermined model, rendering them incapable of capturing the subjects’ full details. Alternatively, high-resolution point clouds can be utilized to represent human subjects. However, such three-dimensional data is computationally expensive to process. In the realm of audio, sound source separation requires prior knowledge of the subjects’ locations. However, it may take unacceptably long for sound source localization algorithms to provide this knowledge, which can still be error-prone, especially with moving objects. These challenges make it difficult for AR systems to produce real-time, high-fidelity representations of human subjects for applications such as AR/VR conferencing that mandate negligible system latency. We present Acuity, a real-time system capable of creating high-fidelity representations of human subjects in a virtual scene both visually and aurally. Acuity isolates subjects from high-resolution input point clouds. It reduces the processing overhead by performing background subtraction at a coarse resolution, then applying the detected bounding boxes to fine-grained point clouds. Meanwhile, Acuity leverages an audiovisual sensor fusion approach to expedite sound source separation. The estimated object location in the visual domain guides the acoustic pipeline to isolate the subjects’ voices without running sound source localization. Our results demonstrate that Acuity can isolate multiple subjects’ high-quality point clouds with a maximum latency of 70 ms and average throughput of over 25 fps, while separating audio in less than 30 ms. We provide the source code of Acuity at: https://github.com/nesl/Acuity. 
    more » « less