Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

D3

Benchmarks

Task NameDataset NameSOTA ResultTrend
Object DetectionD3
Full Score42
35
ClassificationD3
Mean Accuracy76.954
30
Described Object DetectionD3 (Abs)
mAP31.5
16
Described Object DetectionD3 (Pres)
mAP32.9
16
Described Object DetectionD3 (Full)
mAP32.5
16
Described Object DetectionD3 XL
mAP25.4
14
Described Object DetectionD3 L
mAP31.3
14
Described Object DetectionD3 (M)
mAP35.3
14
Described Object DetectionD3 (S)
mAP35.5
14
RegressionD3
Average Relative MSE0.023
11
ClassificationD3 0.15 (test)
Mean Accuracy77.126
10
Diverse Object DetectionD3 (Inter-scenario)
mAP (FULL)5.7
10
Diverse Object DetectionD3 (Intra-scenario)
mAP (FULL)21.6
10
Visual GroundingD3 (Inter-scenario)
APb (Full)2,100
10
Visual GroundingD3 Intra-scenario
APb (Full)37.5
10
Aspect-level sentiment classificationD3
Accuracy81.3
9
Time-Domain PredictionD3
NMSE (dB)-15.52
6
SegmentationD3 trained on D0 (test)
Dice Score94.82
5
SegmentationD3 evaluated after training on D2
Dice94.43
5
Knee cartilage segmentationD3
Dice93.09
5
Reliability AssessmentD3 (test)
AU-ARC94.56
5
Text-to-image generationD3 (test)
LCM Text Alignment0.5236
5
Frequency-Domain PredictionD3
NMSE (dB)-9.03
5
People countingD3 unseen environment full dataset (evaluation)
AP55
4
Underwater Image RestorationD3
ΔE0027.82
4
Showing 25 of 30 rows