Beyond-Hungarian replaces the costly Hungarian assignment in DETR with a differentiable, query‑free matching scheme that halves training latency, boosts large‑object AP by 4.2 points, and introduces a GT‑Probe module and dual‑loss framework, while detailing trade‑offs, ablations, and future challenges.
The article dissects the query‑activation imbalance in DETR‑based detectors, explains PaQ‑DETR’s pattern‑sharing and quality‑aware assignment mechanisms, and shows how these jointly raise detection mAP by up to 4.2% on COCO with less than 5% extra FLOPs.
The article dissects EFSI‑DETR, a UAV small‑object detector that combines simulated frequency processing with dynamic semantic enhancement to overcome pixel scarcity, static fusion, and ignored frequency cues, delivering 188 FPS and a 5.8% APₛ gain on VisDrone while remaining lightweight.
D-FINE introduces Fine-grained Distribution Refinement and Global Optimal Localization Self-Distillation to overhaul DETR's bounding-box regression, achieving 54‑59% AP on COCO and Objects365 at 78‑124 FPS while surpassing YOLO and RT-DETR in both accuracy and speed.
DNTextSpotter is an arbitrary-shaped scene text spotting model using the DETR architecture with an improved denoising training scheme that adds noise to Bézier control points and employs mask‑sliding character queries, achieving significant benchmark gains without extra inference cost and enabling robust text recognition in challenging environments.
This article reviews the NeurIPS 2021 paper introducing BOVText, a large‑scale bilingual video‑text dataset with over 2,000 videos and 1.75 million frames, and describes its transformer‑based end‑to‑end video text spotter that integrates EAST encoding into DETR, covering dataset collection, annotation, architecture, and experimental results.
This article explores how Transformer architectures, originally designed for NLP, are adapted for visual tasks, detailing pioneering models such as DETR, CBAM, NLNet, SENet, and GCNet, and explains their structures, attention mechanisms, advantages, and experimental findings for image processing.
