DRGBT-1K: A Large-scale High-quality Benchmark for Dynamic RGBT Tracking

2026-07-22 · Source: Computer Vision and Pattern Recognition · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Computer Vision · Depth: Expert, quick

Summary

DRGBT-1K is a new large-scale, high-quality benchmark designed for Dynamic RGBT (DRGBT) tracking, which involves continuously localizing targets despite varying sensing modalities and observation platforms over time. Addressing limitations in existing benchmarks that lack real dynamic modality variations and cross-platform transitions, DRGBT-1K comprises 1,045 real-world sequences and 795K RGBT frame pairs. These were collected using UAVs and handheld RGBT devices, capturing diverse scenes, pronounced viewpoint changes, and target appearance discontinuities. The benchmark provides comprehensive annotations, including dense bounding boxes, 24 target categories, more than 15 scene types, 15 challenge attributes, and frame-level modality/platform labels. It also evaluates 20 representative multimodal tracking methods under a unified protocol and offers UGVT-1K for unaligned multimodal tracking, alongside an online evaluation platform and leaderboard.

Key takeaway

For computer vision engineers developing or evaluating RGBT tracking algorithms, DRGBT-1K offers a critical resource. Existing benchmarks fall short in simulating real-world dynamic modality and cross-platform transitions, making DRGBT-1K essential for rigorous testing. Utilize this large-scale, high-quality dataset to validate your tracker's robustness, explore unaligned tracking with UGVT-1K, and contribute to the online leaderboard for comparative analysis.

Key insights

DRGBT-1K provides a robust benchmark for evaluating dynamic RGBT tracking under real-world modality and platform variations.

Principles

• Real-world collaborative perception systems require dynamic modality and cross-platform evaluation.
• Comprehensive annotations are crucial for fine-grained tracker evaluation.

Method

Constructed DRGBT-1K with 1,045 real-world sequences and 795K RGBT frame pairs, annotated for dynamic modality, platform, and target attributes.

In practice

• Evaluate RGBT trackers against DRGBT-1K's diverse real-world scenarios.
• Explore unaligned multimodal tracking using the UGVT-1K dataset.
• Contribute and compare methods on the DRGBT-1K online leaderboard.

Topics

• Dynamic RGBT Tracking
• Multimodal Tracking
• Computer Vision Benchmarks
• UAV-ground Collaborative Tracking
• Dataset Annotation
• Object Tracking

Best for: Research Scientist, AI Scientist, Computer Vision Engineer

Related on AIssential

• DRGBT-1K: A Large-scale High-quality Benchmark for Dynamic RGBT Tracking — Robust dynamic RGBT tracking necessitates benchmarks that accurately reflect real-world modality and platform variations.
• Dual-Correlation Hypergraph Network for Unaligned RGBT Video Object Detection and A Large-scale Benchmark — DHNet and DVT-VOD1000 advance RGBT VOD by resolving spatial misalignment and data scarcity through dual-correlation hypergraphs and a large-scale benchmark.
• T-STAR: A Large-Scale Benchmark for Spatio-Temporal Panoptic Scene Graph Generation in Satellite Video — Spatio-temporal panoptic scene graph generation in satellite video is benchmarked with T-STAR and a unified framework.
• Active Adversarial Perturbation-driven Associative Memory Retrieval for RGB-Event Visual Object Tracking — APRTrack improves RGB-Event tracking robustness by simulating real-world degradations and using a novel historical information retrieval method.
• DF3DV-1K: A Large-Scale Dataset and Benchmark for Distractor-Free Novel View Synthesis — DF3DV-1K is a new dataset for benchmarking distractor-free novel view synthesis methods.
• Flat-Pack Bench: Evaluating Spatio-Temporal Understanding in Large Vision-Language Models through Furniture Assembly — Current LVLMs significantly struggle with fine-grained spatio-temporal reasoning and object interaction in complex video scenarios.

See Counsel's argued verdicts on the open AI decisions leaders are weighing →

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Computer Vision and Pattern Recognition.