rfdetr
Datasets
All datasets matching “rfdetr”rfdetr-segmentation-leibniz-dataset
Dataset Card for Leibniz's Manuscripts (Instance Segmentation Dataset)
This dataset comprises instance segmentation annotations in raw COCO format, used to train an RF-DETR-Seg-nano model for the automatic recognition of textual, graphical, and mathematical expression zones within the manuscripts of the philosopher and mathematician Gottfried Wilhelm Leibniz (17th-early 18th c.).
Dataset Details
Uses
Direct Use
This dataset is… See the full description on the dataset page: https://huggingface.co/datasets/DenisaBumba/rfdetr-segmentation-leibniz-dataset.chess-rfdetr-cocorfdetr-roadsign-agree1-noaug-pr-roc
PR / ROC eval vs human GT — Francesco/road-signs-6ih4y (test)
IoU=0.5, operating threshold=0.5. Predictions matched greedily to human objects GT by class + IoU.
model
Precision
Recall
F1
AP
ROC-AUC
merve/rfdetr-roadsign-agree1-noaug
0.4992
0.5822
0.5375
0.4655
0.9846
ROC note. Detection has no natural true-negative pool, so the ROC treats each prediction as one sample (TP=1 / FP=0) with the model confidence as the score — it measures how well confidence… See the full description on the dataset page: https://huggingface.co/datasets/merve/rfdetr-roadsign-agree1-noaug-pr-roc.rf-detr-keypoint-latency-3090-20260904
rf-detr keypoint inference latency on 3090 (2026-09-04)
Measures the rf-detr-keypoint-preview-xlarge checkpoint (Apache-2.0
upstream, 129 M params, XL preview variant) per-frame forward wall on
the RTX 3090 at 816×816. The number is one input to the RFD 1170
body-presence-half latency budget (SIDEKICK owns the voice-turn half;
this repository is the body-presence half).
The numbers, at a glance
metric
run_1
run_2
mean forward
64.13 ms
63.84 ms
p50… See the full description on the dataset page: https://huggingface.co/datasets/chibifire/rf-detr-keypoint-latency-3090-20260904.rfdetr-roadsign-pr-roc
PR / ROC eval vs human GT — Francesco/road-signs-6ih4y (test)
IoU=0.5, operating threshold=0.5. Predictions matched greedily to human objects GT by class + IoU.
model
Precision
Recall
F1
AP
ROC-AUC
merve/rfdetr-roadsign-agree1-large-noaug
0.4633
0.5728
0.5123
0.4465
0.9864
merve/rfdetr-roadsign-agree2-large-noaug
0.4353
0.4896
0.4609
0.3978
0.9879
ROC note. Detection has no natural true-negative pool, so the ROC treats each prediction as one sample (TP=1 /… See the full description on the dataset page: https://huggingface.co/datasets/merve/rfdetr-roadsign-pr-roc.webui7k-coco-rfdetr