Skip to content
Agientry
← Back to Videos

MetricTrail Estimates Long-Video Pose Without Long Training

Loading video
Loading video

NVIDIA MetricTrail estimates long-video pose without long-sequence training while maintaining metric scale and trajectory consistency, with code and models released.

Pose estimationLong VideoVisual OdometryNVIDIA
Category: perception
Author: @ShengyHuang
Date: 2026-11-05T00:00:00
Duration: 99.0s
Source: https://x.com/ShengyHuang/status/2118416775708881094
Reference: https://research.nvidia.com/labs/sil/projects/metrictrail/

Related Videos

VFlash: Parallel Visual Geometry for Physical AI
0:56

VFlash: Parallel Visual Geometry for Physical AI

@DrJimFan · 30 views · 2026-11-05
Physical AISystem One Models
DirtyMoCap: Robust Motion Capture from Noisy, Unconstrained Markers
0:23

DirtyMoCap: Robust Motion Capture from Noisy, Unconstrained Markers

@yuliangxiu · 289 views · 2026-10-11
Motion CaptureDirtyMoCap
3D Climbing Analysis with iPhone LiDAR
0:29

3D Climbing Analysis with iPhone LiDAR

@jeremyparkphd · 1k views · 2026-09-24
Computer VisionLiDAR
PyCuVSLAM: GPU-Accelerated Visual SLAM for Python
0:34

PyCuVSLAM: GPU-Accelerated Visual SLAM for Python

@tokufxug · 613 views · 2026-09-21
Visual SLAMcuVSLAM
Google Gemini agentic video understanding
0:24

Google Gemini agentic video understanding

@Gorden_Sun · 714 views · 2026-09-02
GoogleGemini
NanoOWL: Zero-Shot Real-Time Object Detection by NVIDIA
0:14

NanoOWL: Zero-Shot Real-Time Object Detection by NVIDIA

@simplifyinAI · 567 views · 2026-08-24
NVIDIANanoOWL

As an Amazon Associate, we earn from qualifying purchases.