FPG-NAS: FLOPs-Aware Gated Differentiable Neural Architecture Search for Efficient 6DoF Pose Estimation

2025-08-05

AI Summary

Efficient 6DoF object pose estimation under resource constraints remains challenging. Method: This paper introduces, for the first time, differentiable Neural Architecture Search (NAS) to this task, proposing a FLOPs-aware gated differentiable NAS framework. It constructs a task-specific search space and designs a gated selection mechanism over discrete multi-candidate operators, enabling exploration of an ultra-large architecture space (~10⁹²) while incorporating FLOPs regularization for end-to-end joint optimization of accuracy and computational efficiency. Contribution/Results: On LINEMOD and SPEED+, the searched lightweight models significantly outperform state-of-the-art methods under strict FLOPs budgets, achieving superior trade-offs between pose estimation accuracy and inference speed. This demonstrates the effectiveness and generalizability of task-driven neural architecture search for efficient 6DoF pose estimation.

Technology Category

• Search and Optimization: Learning to Search
• Machine Learning: Deep Neural Architectures and Foundation Models
• Intelligent Robots: State Estimation

Application Category

• Search and Retrieval-Augmented AI: Efficiency and scalability of Web search engines
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

We introduce FPG-NAS, a FLOPs-aware Gated Differentiable Neural Architecture Search framework for efficient 6DoF object pose estimation. Estimating 3D rotation and translation from a single image has been widely investigated yet remains computationally demanding, limiting applicability in resource-constrained scenarios. FPG-NAS addresses this by proposing a specialized differentiable NAS approach for 6DoF pose estimation, featuring a task-specific search space and a differentiable gating mechanism that enables discrete multi-candidate operator selection, thus improving architectural diversity. Additionally, a FLOPs regularization term ensures a balanced trade-off between accuracy and efficiency. The framework explores a vast search space of approximately 10 extsuperscript{92} possible architectures. Experiments on the LINEMOD and SPEED+ datasets demonstrate that FPG-NAS-derived models outperform previous methods under strict FLOPs constraints. To the best of our knowledge, FPG-NAS is the first differentiable NAS framework specifically designed for 6DoF object pose estimation.

Problem

• Efficient 6DoF pose estimation under FLOPs constraints
• Balancing accuracy and computational efficiency in NAS
• Specialized differentiable NAS for diverse architecture selection

Innovation

• FLOPs-aware gated differentiable NAS framework
• Task-specific search space with operator selection
• FLOPs regularization for accuracy-efficiency balance