Farm3D: Learning Articulated 3D Animals by Distilling 2D Diffusion

• URL: http://arxiv.org/abs/2304.10535v3
• Date: Tue, 14 May 2024 14:37:32 GMT
• Title: Farm3D: Learning Articulated 3D Animals by Distilling 2D Diffusion
• Authors: Tomas Jakab, Ruining Li, Shangzhe Wu, Christian Rupprecht, Andrea Vedaldi,
• Abstract summary: We present Farm3D, a method for learning category-specific 3D reconstructors for articulated objects. We propose a framework that uses an image generator, such as Stable Diffusion, to generate synthetic training data. Our network can be used for analysis, including monocular reconstruction, or for synthesis, generating articulated assets for real-time applications such as video games.
• Score: 67.71624118802411
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: We present Farm3D, a method for learning category-specific 3D reconstructors for articulated objects, relying solely on "free" virtual supervision from a pre-trained 2D diffusion-based image generator. Recent approaches can learn a monocular network that predicts the 3D shape, albedo, illumination, and viewpoint of any object occurrence, given a collection of single-view images of an object category. However, these approaches heavily rely on manually curated clean training data, which are expensive to obtain. We propose a framework that uses an image generator, such as Stable Diffusion, to generate synthetic training data that are sufficiently clean and do not require further manual curation, enabling the learning of such a reconstruction network from scratch. Additionally, we incorporate the diffusion model as a score to enhance the learning process. The idea involves randomizing certain aspects of the reconstruction, such as viewpoint and illumination, generating virtual views of the reconstructed 3D object, and allowing the 2D network to assess the quality of the resulting image, thus providing feedback to the reconstructor. Unlike work based on distillation, which produces a single 3D asset for each textual prompt, our approach yields a monocular reconstruction network capable of outputting a controllable 3D asset from any given image, whether real or generated, in a single forward pass in a matter of seconds. Our network can be used for analysis, including monocular reconstruction, or for synthesis, generating articulated assets for real-time applications such as video games.

Related papers

• DistillNeRF: Perceiving 3D Scenes from Single-Glance Images by Distilling Neural Fields and Foundation Model Features [65.8738034806085]
  DistillNeRF is a self-supervised learning framework for understanding 3D environments in autonomous driving. It predicts a rich neural scene representation from sparse, single-frame multi-view camera inputs. It is trained self-supervised with differentiable rendering to reconstruct RGB, depth, or feature images.
  arXiv  Detail & Related papers  (2024-06-17T21:15:13Z)
• Viewset Diffusion: (0-)Image-Conditioned 3D Generative Models from 2D Data [76.38261311948649]
  Viewset Diffusion is a diffusion-based generator that outputs 3D objects while only using multi-view 2D data for supervision. We train a diffusion model to generate viewsets, but design the neural network generator to reconstruct internally corresponding 3D models. The model performs reconstruction efficiently, in a feed-forward manner, and is trained using only rendering losses using as few as three views per viewset.
  arXiv  Detail & Related papers  (2023-06-13T16:18:51Z)
• Anything-3D: Towards Single-view Anything Reconstruction in the Wild [61.090129285205805]
  We introduce Anything-3D, a methodical framework that ingeniously combines a series of visual-language models and the Segment-Anything object segmentation model. Our approach employs a BLIP model to generate textural descriptions, utilize the Segment-Anything model for the effective extraction of objects of interest, and leverages a text-to-image diffusion model to lift object into a neural radiance field.
  arXiv  Detail & Related papers  (2023-04-19T16:39:51Z)
• 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)
• 2D GANs Meet Unsupervised Single-view 3D Reconstruction [21.93671761497348]
  controllable image generation based on pre-trained GANs can benefit a wide range of computer vision tasks. We propose a novel image-conditioned neural implicit field, which can leverage 2D supervisions from GAN-generated multi-view images. The effectiveness of our approach is demonstrated through superior single-view 3D reconstruction results of generic objects.
  arXiv  Detail & Related papers  (2022-07-20T20:24:07Z)
• Style Agnostic 3D Reconstruction via Adversarial Style Transfer [23.304453155586312]
  Reconstructing the 3D geometry of an object from an image is a major challenge in computer vision. We propose an approach that enables a differentiable-based learning of 3D objects from images with backgrounds.
  arXiv  Detail & Related papers  (2021-10-20T21:24:44Z)
• Model-based 3D Hand Reconstruction via Self-Supervised Learning [72.0817813032385]
  Reconstructing a 3D hand from a single-view RGB image is challenging due to various hand configurations and depth ambiguity. We propose S2HAND, a self-supervised 3D hand reconstruction network that can jointly estimate pose, shape, texture, and the camera viewpoint. For the first time, we demonstrate the feasibility of training an accurate 3D hand reconstruction network without relying on manual annotations.
  arXiv  Detail & Related papers  (2021-03-22T10:12:43Z)
• An Effective Loss Function for Generating 3D Models from Single 2D Image without Rendering [0.0]
  Differentiable rendering is a very successful technique that applies to a Single-View 3D Reconstruction. Currents use losses based on pixels between a rendered image of some 3D reconstructed object and ground-truth images from given matched viewpoints to optimise parameters of the 3D shape. We propose a novel effective loss function that evaluates how well the projections of reconstructed 3D point clouds cover the ground truth object's silhouette.
  arXiv  Detail & Related papers  (2021-03-05T00:02:18Z)
• From Image Collections to Point Clouds with Self-supervised Shape and Pose Networks [53.71440550507745]
  Reconstructing 3D models from 2D images is one of the fundamental problems in computer vision. We propose a deep learning technique for 3D object reconstruction from a single image. We learn both 3D point cloud reconstruction and pose estimation networks in a self-supervised manner.
  arXiv  Detail & Related papers  (2020-05-05T04:25:16Z)

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.