Curiosity-driven 3D Scene Structure from Single-image Self-supervision

• URL: http://arxiv.org/abs/2012.01230v2
• Date: Fri, 19 Feb 2021 13:55:40 GMT
• Title: Curiosity-driven 3D Scene Structure from Single-image Self-supervision
• Authors: David Griffiths, Jan Boehm, Tobias Ritschel
• Abstract summary: Previous work has demonstrated learning isolated 3D objects from 2D-only self-supervision. Here we set out to extend this to entire 3D scenes made out of multiple objects, including their location, orientation and type. The resulting system converts 2D images of different virtual or real images into complete 3D scenes, learned only from 2D images of those scenes.
• Score: 22.527696847086574
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Previous work has demonstrated learning isolated 3D objects (voxel grids, point clouds, meshes, etc.) from 2D-only self-supervision. Here we set out to extend this to entire 3D scenes made out of multiple objects, including their location, orientation and type, and the scenes illumination. Once learned, we can map arbitrary 2D images to 3D scene structure. We analyze why analysis-by-synthesis-like losses for supervision of 3D scene structure using differentiable rendering is not practical, as it almost always gets stuck in local minima of visual ambiguities. This can be overcome by a novel form of training: we use an additional network to steer the optimization itself to explore the full gamut of possible solutions \ie to be curious, and hence, to resolve those ambiguities and find workable minima. The resulting system converts 2D images of different virtual or real images into complete 3D scenes, learned only from 2D images of those scenes.

Related papers

• Lift3D: Zero-Shot Lifting of Any 2D Vision Model to 3D [95.14469865815768]
  2D vision models can be used for semantic segmentation, style transfer or scene editing, enabled by large-scale 2D image datasets. However, extending a single 2D vision operator like scene editing to 3D typically requires a highly creative method specialized to that task. In this paper, we propose Lift3D, which trains to predict unseen views on feature spaces generated by a few visual models. We even outperform state-of-the-art methods specialized for the task in question.
  arXiv  Detail & Related papers  (2024-03-27T18:13:16Z)
• BUOL: A Bottom-Up Framework with Occupancy-aware Lifting for Panoptic 3D Scene Reconstruction From A Single Image [33.126045619754365]
  BUOL is a framework with Occupancy-aware Lifting to address the two issues for panoptic 3D scene reconstruction from a single image. Our method shows a tremendous performance advantage over state-of-the-art methods on synthetic dataset 3D-Front and real-world dataset Matterport3D.
  arXiv  Detail & Related papers  (2023-06-01T17:56:49Z)
• CLIP$^2$: Contrastive Language-Image-Point Pretraining from Real-World Point Cloud Data [80.42480679542697]
  We propose Contrastive Language-Image-Point Cloud Pretraining (CLIP$2$) to learn the transferable 3D point cloud representation in realistic scenarios. Specifically, we exploit naturally-existed correspondences in 2D and 3D scenarios, and build well-aligned and instance-based text-image-point proxies from those complex scenarios.
  arXiv  Detail & Related papers  (2023-03-22T09:32:45Z)
• 3inGAN: Learning a 3D Generative Model from Images of a Self-similar Scene [34.2144933185175]
  3inGAN is an unconditional 3D generative model trained from 2D images of a single self-similar 3D scene. We show results on semi-stochastic scenes of varying scale and complexity, obtained from real and synthetic sources.
  arXiv  Detail & Related papers  (2022-11-27T18:03:21Z)
• Learning 3D Scene Priors with 2D Supervision [37.79852635415233]
  We propose a new method to learn 3D scene priors of layout and shape without requiring any 3D ground truth. Our method represents a 3D scene as a latent vector, from which we can progressively decode to a sequence of objects characterized by their class categories. Experiments on 3D-FRONT and ScanNet show that our method outperforms state of the art in single-view reconstruction.
  arXiv  Detail & Related papers  (2022-11-25T15:03:32Z)
• Can We Solve 3D Vision Tasks Starting from A 2D Vision Transformer? [111.11502241431286]
  Vision Transformers (ViTs) have proven to be effective in solving 2D image understanding tasks. ViTs for 2D and 3D tasks have so far adopted vastly different architecture designs that are hardly transferable. This paper demonstrates the appealing promise to understand the 3D visual world, using a standard 2D ViT architecture.
  arXiv  Detail & Related papers  (2022-09-15T03:34:58Z)
• Neural Feature Fusion Fields: 3D Distillation of Self-Supervised 2D Image Representations [92.88108411154255]
  We present a method that improves dense 2D image feature extractors when the latter are applied to the analysis of multiple images reconstructible as a 3D scene. We show that our method not only enables semantic understanding in the context of scene-specific neural fields without the use of manual labels, but also consistently improves over the self-supervised 2D baselines.
  arXiv  Detail & Related papers  (2022-09-07T23:24:09Z)
• Recognizing Scenes from Novel Viewpoints [99.90914180489456]
  Humans can perceive scenes in 3D from a handful of 2D views. For AI agents, the ability to recognize a scene from any viewpoint given only a few images enables them to efficiently interact with the scene and its objects. We propose a model which takes as input a few RGB images of a new scene and recognizes the scene from novel viewpoints by segmenting it into semantic categories.
  arXiv  Detail & Related papers  (2021-12-02T18:59:40Z)
• SAT: 2D Semantics Assisted Training for 3D Visual Grounding [95.84637054325039]
  3D visual grounding aims at grounding a natural language description about a 3D scene, usually represented in the form of 3D point clouds, to the targeted object region. Point clouds are sparse, noisy, and contain limited semantic information compared with 2D images. We propose 2D Semantics Assisted Training (SAT) that utilizes 2D image semantics in the training stage to ease point-cloud-language joint representation learning.
  arXiv  Detail & Related papers  (2021-05-24T17:58:36Z)
• Disentangling 3D Prototypical Networks For Few-Shot Concept Learning [29.02523358573336]
  We present neural architectures that disentangle RGB-D images into objects' shapes and styles and a map of the background scene. Our networks incorporate architectural biases that reflect the image formation process, 3D geometry of the world scene, and shape-style interplay.
  arXiv  Detail & Related papers  (2020-11-06T14:08:27Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.