arXiv:2506.00871v3 Announce Type: replace-cross 
Abstract: Predicting accurate future trajectories of pedestrians is essential for autonomous systems but remains a challenging task due to the need for adaptability in different environments and domains. A common approach involves collecting scenario-specific data and performing fine-tuning via backpropagation. However, the need to fine-tune for each new scenario is often impractical for deployment on edge devices. To address this challenge, we introduce TrajICL, an In-Context Learning (ICL) framework for pedestrian trajectory prediction that enables adaptation without fine-tuning on the scenario-specific data at inference time without requiring weight updates. We propose a spatio-temporal similarity-based example selection (STES) method that selects relevant examples from previously observed trajectories within the same scene by identifying similar motion patterns at corresponding locations. To further refine this selection, we introduce prediction-guided example selection (PG-ES), which selects examples based on both the past trajectory and the predicted future trajectory, rather than relying solely on the past trajectory. This approach allows the model to account for long-term dynamics when selecting examples. Finally, instead of relying on small real-world datasets with limited scenario diversity, we train our model on a large-scale synthetic dataset to enhance its prediction ability by leveraging in-context examples. Extensive experiments demonstrate that TrajICL achieves remarkable adaptation across both in-domain and cross-domain scenarios, outperforming even fine-tuned approaches across multiple public benchmarks. Project Page: https://fujiry0.github.io/TrajICL-project-page/.

تسلط دراسة حديثة الضوء على التقدم في توقع مسارات المشاة، وهو أمر حاسم لتطوير الأنظمة المستقلة. تتناول الأبحاث التحديات المتعلقة بالتكيف مع بيئات مختلفة دون الحاجة إلى ضبط شامل، مما يجعلها أكثر عملية للتطبيقات في العالم الحقيقي. يمكن أن تعزز هذه الابتكارات بشكل كبير من كفاءة وموثوقية التقنيات المستقلة، مما يمهد الطريق لملاحة أكثر أمانًا وفعالية في البيئات المعقدة.

Un estudio reciente destaca los avances en la predicción de trayectorias de peatones, lo cual es crucial para el desarrollo de sistemas autónomos. La investigación aborda los desafíos de adaptarse a diversos entornos sin necesidad de un ajuste extenso, lo que lo hace más práctico para aplicaciones del mundo real. Esta innovación podría mejorar significativamente la eficiencia y la fiabilidad de las tecnologías autónomas, allanando el camino para una navegación más segura y efectiva en entornos complejos.

Une étude récente met en lumière les avancées dans la prédiction des trajectoires des piétons, ce qui est crucial pour le développement des systèmes autonomes. La recherche aborde les défis d'adaptation à divers environnements sans nécessiter un réglage approfondi, ce qui la rend plus pratique pour les applications réelles. Cette innovation pourrait considérablement améliorer l'efficacité et la fiabilité des technologies autonomes, ouvrant la voie à une navigation plus sûre et plus efficace dans des contextes complexes.

A recent study highlights advancements in predicting pedestrian trajectories, which is crucial for the development of autonomous systems. The research addresses the challenges of adapting to various environments without the need for extensive fine-tuning, making it more practical for real-world applications. This innovation could significantly enhance the efficiency and reliability of autonomous technologies, paving the way for safer and more effective navigation in complex settings.

Towards Predicting Any Human Trajectory In Context

arXiv:2601.08303v1 Announce Type: new 
Abstract: Recent advances in diffusion transformers (DiTs) have set new standards in image generation, yet remain impractical for on-device deployment due to their high computational and memory costs. In this work, we present an efficient DiT framework tailored for mobile and edge devices that achieves transformer-level generation quality under strict resource constraints. Our design combines three key components. First, we propose a compact DiT architecture with an adaptive global-local sparse attention mechanism that balances global context modeling and local detail preservation. Second, we propose an elastic training framework that jointly optimizes sub-DiTs of varying capacities within a unified supernetwork, allowing a single model to dynamically adjust for efficient inference across different hardware. Finally, we develop Knowledge-Guided Distribution Matching Distillation, a step-distillation pipeline that integrates the DMD objective with knowledge transfer from few-step teacher models, producing high-fidelity and low-latency generation (e.g., 4-step) suitable for real-time on-device use. Together, these contributions enable scalable, efficient, and high-quality diffusion models for deployment on diverse hardware.

قدمت SnapGen++ إطار عمل جديد يستخدم المحولات الانتشارية (DiTs) لتمكين توليد صور عالية الدقة بكفاءة على الأجهزة المحمولة وأجهزة الحافة، مما يعالج التكاليف العالية للحوسبة والذاكرة التي أعاقت نشرها على الأجهزة.

SnapGen++ ha presentado un nuevo marco que utiliza transformadores de difusión (DiTs) para permitir la generación eficiente de imágenes de alta fidelidad en dispositivos móviles y de borde, abordando los altos costos computacionales y de memoria que han obstaculizado el despliegue en dispositivos.

SnapGen++ a introduit un nouveau cadre utilisant des transformateurs de diffusion (DiTs) pour permettre une génération d'images haute fidélité efficace sur des appareils mobiles et en périphérie, répondant aux coûts computationnels et de mémoire élevés qui ont entravé le déploiement sur appareil.

SnapGen++ has introduced a new framework leveraging diffusion transformers (DiTs) to enable efficient high-fidelity image generation on mobile and edge devices, addressing the high computational and memory costs that have hindered on-device deployment.

SnapGen++: Unleashing Diffusion Transformers for Efficient High-Fidelity Image Generation on Edge Devices

One More Thing in AI – Your Shortcut to AI Mastery

Towards Predicting Any Human Trajectory In Context

Was this article worth reading? Share it

One More Thing in AI

LucidQuery AI

GPTHumanizer

Attentive AI

Portfolio Backtest

Bypass

Ready to build your own newsroom?