Dreamweaver: Learning Compositional World Representations from Pixels

• URL: http://arxiv.org/abs/2501.14174v2
• Date: Tue, 18 Feb 2025 08:16:03 GMT
• Title: Dreamweaver: Learning Compositional World Representations from Pixels
• Authors: Junyeob Baek, Yi-Fu Wu, Gautam Singh, Sungjin Ahn,
• Abstract summary: Humans have an innate ability to decompose their perceptions of the world into objects and their attributes. This cognitive process enables us to imagine novel futures by recombining familiar concepts. We propose a neural architecture designed to discover hierarchical and compositional representations from raw videos.
• Score: 22.978369848454616
• License:
• Abstract: Humans have an innate ability to decompose their perceptions of the world into objects and their attributes, such as colors, shapes, and movement patterns. This cognitive process enables us to imagine novel futures by recombining familiar concepts. However, replicating this ability in artificial intelligence systems has proven challenging, particularly when it comes to modeling videos into compositional concepts and generating unseen, recomposed futures without relying on auxiliary data, such as text, masks, or bounding boxes. In this paper, we propose Dreamweaver, a neural architecture designed to discover hierarchical and compositional representations from raw videos and generate compositional future simulations. Our approach leverages a novel Recurrent Block-Slot Unit (RBSU) to decompose videos into their constituent objects and attributes. In addition, Dreamweaver uses a multi-future-frame prediction objective to capture disentangled representations for dynamic concepts more effectively as well as static concepts. In experiments, we demonstrate our model outperforms current state-of-the-art baselines for world modeling when evaluated under the DCI framework across multiple datasets. Furthermore, we show how the modularized concept representations of our model enable compositional imagination, allowing the generation of novel videos by recombining attributes from different objects.

Related papers

• Visual Representation Learning with Stochastic Frame Prediction [90.99577838303297]
  This paper revisits the idea of video generation that learns to capture uncertainty in frame prediction. We design a framework that trains a frame prediction model to learn temporal information between frames. We find this architecture allows for combining both objectives in a synergistic and compute-efficient manner.
  arXiv  Detail & Related papers  (2024-06-11T16:05:15Z)
• DreamCreature: Crafting Photorealistic Virtual Creatures from Imagination [140.1641573781066]
  We introduce a novel task, Virtual Creatures Generation: Given a set of unlabeled images of the target concepts, we aim to train a T2I model capable of creating new, hybrid concepts. We propose a new method called DreamCreature, which identifies and extracts the underlying sub-concepts. The T2I thus adapts to generate novel concepts with faithful structures and photorealistic appearance.
  arXiv  Detail & Related papers  (2023-11-27T01:24:31Z)
• OC-NMN: Object-centric Compositional Neural Module Network for Generative Visual Analogical Reasoning [49.12350554270196]
  We show how modularity can be leveraged to derive a compositional data augmentation framework inspired by imagination. Our method, denoted Object-centric Compositional Neural Module Network (OC-NMN), decomposes visual generative reasoning tasks into a series of primitives applied to objects without using a domain-specific language.
  arXiv  Detail & Related papers  (2023-10-28T20:12:58Z)
• Compositional diversity in visual concept learning [18.907108368038216]
  Humans leverage compositionality to efficiently learn new concepts, understanding how familiar parts can combine together to form novel objects. Here, we study how people classify and generate alien figures'' with rich relational structure. We develop a Bayesian program induction model which searches for the best programs for generating the candidate visual figures.
  arXiv  Detail & Related papers  (2023-05-30T19:30:50Z)
• Recursive Neural Programs: Variational Learning of Image Grammars and Part-Whole Hierarchies [1.5990720051907859]
  We introduce Recursive Neural Programs (RNPs) to address the part-whole hierarchy learning problem. RNPs are the first neural generative model to address the part-whole hierarchy learning problem. Our results show that RNPs provide an intuitive and explainable way of composing objects and scenes.
  arXiv  Detail & Related papers  (2022-06-16T22:02:06Z)
• AIGenC: An AI generalisation model via creativity [1.933681537640272]
  Inspired by cognitive theories of creativity, this paper introduces a computational model (AIGenC) It lays down the necessary components to enable artificial agents to learn, use and generate transferable representations. We discuss the model's capability to yield better out-of-distribution generalisation in artificial agents.
  arXiv  Detail & Related papers  (2022-05-19T17:43:31Z)
• Learning Multi-Object Dynamics with Compositional Neural Radiance Fields [63.424469458529906]
  We present a method to learn compositional predictive models from image observations based on implicit object encoders, Neural Radiance Fields (NeRFs), and graph neural networks. NeRFs have become a popular choice for representing scenes due to their strong 3D prior. For planning, we utilize RRTs in the learned latent space, where we can exploit our model and the implicit object encoder to make sampling the latent space informative and more efficient.
  arXiv  Detail & Related papers  (2022-02-24T01:31:29Z)
• 3D Neural Scene Representations for Visuomotor Control [78.79583457239836]
  We learn models for dynamic 3D scenes purely from 2D visual observations. A dynamics model, constructed over the learned representation space, enables visuomotor control for challenging manipulation tasks.
  arXiv  Detail & Related papers  (2021-07-08T17:49:37Z)
• SketchEmbedNet: Learning Novel Concepts by Imitating Drawings [125.45799722437478]
  We explore properties of image representations learned by training a model to produce sketches of images. We show that this generative, class-agnostic model produces informative embeddings of images from novel examples, classes, and even novel datasets in a few-shot setting.
  arXiv  Detail & Related papers  (2020-08-27T16:43:28Z)

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.