Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

PhotoAgent: A Robotic Photographer with Spatial and Aesthetic Understanding

Lirong Che, Zhenfeng Gan, Yanbo Chen, Junbo Tan, Xueqian Wang

Abstract

Embodied agents for creative tasks like photography must bridge the semantic gap between high-level language commands and geometric control. We introduce PhotoAgent, an agent that achieves this by integrating Large Multimodal Models (LMMs) reasoning with a novel control paradigm. PhotoAgent first translates subjective aesthetic goals into solvable geometric constraints via LMM-driven, chain-of-thought (CoT) reasoning, allowing an analytical solver to compute a high-quality initial viewpoint. This initial pose is then iteratively refined through visual reflection within a photorealistic internal world model built with 3D Gaussian Splatting (3DGS). This ``mental simulation'' replaces costly and slow physical trial-and-error, enabling rapid convergence to aesthetically superior results. Evaluations confirm that PhotoAgent excels in spatial reasoning and achieves superior final image quality.

PhotoAgent: A Robotic Photographer with Spatial and Aesthetic Understanding

Abstract

Embodied agents for creative tasks like photography must bridge the semantic gap between high-level language commands and geometric control. We introduce PhotoAgent, an agent that achieves this by integrating Large Multimodal Models (LMMs) reasoning with a novel control paradigm. PhotoAgent first translates subjective aesthetic goals into solvable geometric constraints via LMM-driven, chain-of-thought (CoT) reasoning, allowing an analytical solver to compute a high-quality initial viewpoint. This initial pose is then iteratively refined through visual reflection within a photorealistic internal world model built with 3D Gaussian Splatting (3DGS). This ``mental simulation'' replaces costly and slow physical trial-and-error, enabling rapid convergence to aesthetically superior results. Evaluations confirm that PhotoAgent excels in spatial reasoning and achieves superior final image quality.
Paper Structure (22 sections, 9 equations, 7 figures, 3 tables, 1 algorithm)

This paper contains 22 sections, 9 equations, 7 figures, 3 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Related Work
  3. Robotic Photography
  4. Reasoning Architectures for Embodied Agents
  5. World Models for Visual Foresight
  6. Method
  7. System Overview
  8. Intention Parsing
  9. Geometric Solving
  10. Reflective Reasoning
  11. Experiments
  12. Spatial Reasoning Evaluation
  13. Experimental Setup
  14. Baseline
  15. Comparative Experiments
  16. ...and 7 more sections

Figures (7)

  • Figure 1: Overview of PhotoAgent and its capabilities. (a) illustrates the inefficiency of real-world trial-and-error, while PhotoAgent leverages internal simulation to achieve one-shot success. (b) highlights three key capabilities.
  • Figure 2: Overall cognitive architecture of PhotoAgent.
  • Figure 3: Intention Parsing workflow demonstration. "Take a photo of the toys with visual tension, like the reference."
  • Figure 4: Closed-loop reflective reasoning. Starting from an internal 3DGS “imagined world”, the agent renders candidate views, critiques them via the LMM, and issues optimized motion commands.
  • Figure 5: Performances of our method and baselines on three tasks of different levels.
  • ...and 2 more figures