DSFlash：实时全景场景图生成的全方位方案
DSFlash: Comprehensive Panoptic Scene Graph Generation in Realtime
对需要把场景图生成部署到边缘设备或视频流的研究者，DSFlash 提供了一个低延迟、可单卡训练的实时全景场景图生成基线，其「不牺牲性能换速度」且输出完整（非仅显著）关系的设定，可直接作为下游具身推理的中间表示模块复用。
原文依据 smaller DSFlash variants achieving 56 frames per second on a standard RTX 3090 GPU
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论场景图生成（SGG）旨在从图像中提取详细的图结构，这种表示作为复杂下游任务（如具身智能体推理）的稳健中间步骤具有重要前景。然而，在真实世界应用中——尤其是在资源受限的边缘设备上——实际部署需要速度与资源效率，而这些挑战在现有研究中受到的关注有限。为弥合这一差距，我们提出 DSFlash，一种面向全景场景图生成的低延迟模型，旨在克服上述局限。DSFlash 能够高效处理视频流，且性能不逊于现有最优方法，其中较小的 DSFlash 变体在标准 RTX 3090 GPU 上可达每秒 56 帧。关键在于，与以往常局限于显著关系的方法不同，DSFlash 计算全面的场景图，在保持优越延迟的同时提供更丰富的上下文信息。此外，DSFlash 对资源要求较低，在单块已有九年历史的 GTX 1080 GPU 上训练耗时不足 24 小时。这种可及性使 DSFlash 特别适合计算资源有限的研究者与从业者，使他们能够针对专门应用调整和微调 SGG 模型。
英文原文摘要 1356 字符
Scene Graph Generation (SGG) aims to extract a detailed graph structure from an image, a representation that holds significant promise as a robust intermediate step for complex downstream tasks like reasoning for embodied agents. However, practical deployment in real-world applications - especially on resource constrained edge devices - requires speed and resource efficiency, challenges that have received limited attention in existing research. To bridge this gap, we introduce DSFlash, a low-latency model for panoptic scene graph generation designed to overcome these limitations. DSFlash can efficiently process video streams without compromising performance against existing state-of-the-art methods, with smaller DSFlash variants achieving 56 frames per second on a standard RTX 3090 GPU. Crucially, unlike prior approaches that often restrict themselves to salient relationships, DSFlash computes comprehensive scene graphs, offering richer contextual information while maintaining its superior latency. Furthermore, DSFlash is light on resources, requiring less than 24 hours to train on a single, nine-year-old GTX 1080 GPU. This accessibility makes DSFlash particularly well-suited for researchers and practitioners operating with limited computational resources, empowering them to adapt and fine-tune SGG models for specialized applications.