Project AURIONAI Use-Cases Research & Innovation
University of Science and Technology of China (USTC)
vision

D-FINE

Real-time DETR detector from USTC; RT-DETR lineage is unverified.

University of Science and Technology of China (USTC)University of Science and Technology of China (USTC)ChinaInstitute of Artificial Intelligence, Hefei Comprehensive National Science CenterInstitute of Artificial Intelligence, Hefei Comprehensive National Science CenterChina
Figure from the D-FINE paper: Comparisons with other detectors in terms of latency (left), model size (mid), and computational cost (right).
Levelmodel
Classvision
Parameters62.9M
Contextnot in the notes
Released17 Oct 2024
Opennessopen weights permissive
Runs onsingle datacentre GPU
Statusactive

Benchmark Results

independent
COCO val valdomain
val, M size, as reported in RF-DETR paperindependent
55
03570

Score in AP. Higher is better.

Nov 2025 arxiv.org
val, S size, as reported in RF-DETR paperindependent
50.6
03570

Score in AP. Higher is better.

Nov 2025 arxiv.org
val, N size, as reported in RF-DETR paperindependent
42.7
03570

Score in AP. Higher is better.

Nov 2025 arxiv.org
Where this comes from
Every score names who ran it: vendor-reported, maintainer-run or independent. A rank is shown with the population it was ranked in (gap rule 3).

Lineage

successor
RT-DETR to D-FINE: successor, low confidenceD-FINE to DEIM: successor, medium confidenceRT-DETRRT-DETRBaidu, Apr 2023D-FINED-FINEUniversity of Science and Technol…DEIMDEIMIntellindust AI Lab, Dec 2024
BaiduRT-DETR is the parent (successor)low confidencenotes
Intellindust AI LabDEIM is a child (successor)medium confidencenotes
Where this comes from
Parent and child relations from model cards, Hugging Face base_model fields and the genealogy tables of the research notes. The tree shows two generations each way, 3 models. Confidence and evidence are the notes' own.

R&D Topics

Real-time object detection (YOLO and DETR lines)exemplar

Identity

Developer
University of Science and Technology of China (USTC), Institute of Artificial Intelligence, Hefei Comprehensive National Science Center
Family
not in the notes
Released
17 Oct 2024 sourcesannouncement page
Weights published
–
Record checked
10 Oct 2026
Where this comes from
Developer and dates from the research notes. Release date is the vendor's public announcement; weights date is when open weights were published.

Architecture and Size

Architecture
dense transformer Real-time DETR
Total parameters
62.9M sources
Active parameters
–
Checkpoint parameters
–
Note
62.9M for dfine-xlarge-coco
Context length
not in the notes
Input
image
Output
image, text
Where this comes from
Parameter counts as the developer states them. Where the checkpoint holds a different total, both are shown (register A16).

Openness and Access

Openness
open weights permissive
Licence
Apache License 2.0 sources
Access routes
open weights download
Where this comes from
The licence file governs over registry tags and aggregator labels.

Deployment

Runs on
single datacentre GPU
Footprint
not in the notes
Runtimes
not in the notes
Where this comes from
Deployment tiers and runtimes as named in the notes. Footprint is the notes' figure, not a measurement by AURION.

Indonesia

Indonesian support
not stated
Regional languages
not in the notes
Jakarta hosting
unknown
Languages
not in the notes
Where this comes from
Indonesia fields are carried on every model; unknown is an allowed value.

Industry Use Cases

No industry use case links this model.

Sources

  1. D-FINE: Redefine Regression Task in DETRs as Fine-grained Distribution Refinement arxiv.org https://arxiv.org/abs/2410.13842
  2. huggingface.co/api/models/ustc-community/dfine-xlarge-coco https://huggingface.co/api/models/ustc-community/dfine-xlarge-coco
  3. RF-DETR: Neural Architecture Search for Real-Time Detection Transformers arxiv.org https://arxiv.org/html/2511.09554v2

To cite this page: Project AURION, AI Use-Cases Research & Innovation, "D-FINE", https://aurion.semai.id/models/ustc-d-fine/, records of October 2026.