List your company
ResearchFormerly WizWand

Object Detection on Pascal VOC (test)

87.5mAPCoLin
53.80462.55271.380.048Sep 26, 2019Nov 10, 2020Dec 26, 2021Feb 11, 2023Mar 28, 2024May 13, 2025Jun 29, 2026
Updated 4mo ago

Evaluation results

MethodLinks
CoLinBackbone=Swin-L, Architecture=RetinaNet, Trained Params=2.39 M, Percentage of trainable backbone parameters=1.23 %, Extra Structure=true2026.0287.5——3.8
LORANDBackbone=Swin-L, Architecture=RetinaNet, Trained Params=5.20 M, Percentage of trainable backbone parameters=2.62 %, Extra Structure=true2026.0286.9——3.2
ADAPTERBackbone=Swin-L, Architecture=RetinaNet, Trained Params=4.61 M, Percentage of trainable backbone parameters=2.33 %, Extra Structure=true2026.0286.7——3
ADAPTFORMERBackbone=Swin-L, Architecture=RetinaNet, Trained Params=2.34 M, Percentage of trainable backbone parameters=1.18 %, Extra Structure=true2026.0286.6——2.9
NORMTUNINGBackbone=Swin-L, Architecture=RetinaNet, Trained Params=0.10 M, Percentage of trainable backbone parameters=0.05 %, Extra Structure=false2026.0285.5——1.8
PARTIAL-1Backbone=Swin-L, Architecture=RetinaNet, Trained Params=28.77 M, Percentage of trainable backbone parameters=14.53 %, Extra Structure=false2026.0285.5——1.8
BITFITBackbone=Swin-L, Architecture=RetinaNet, Trained Params=0.30 M, Percentage of trainable backbone parameters=0.15 %, Extra Structure=false2026.0285.4——1.7
LORABackbone=Swin-L, Architecture=RetinaNet, Trained Params=4.57 M, Percentage of trainable backbone parameters=2.31 %, Extra Structure=true2026.0285.4——1.7
FIXEDBackbone=Swin-L, Architecture=RetinaNet, Trained Params=0.00 M, Percentage of trainable backbone parameters=0.00 %, Extra Structure=false2026.0283.8——0.1
FULLBackbone=Swin-L, Architecture=RetinaNet, Trained Params=198.58 M, Percentage of trainable backbone parameters=100.00 %, Extra Structure=false2026.0283.7———
FR-DETRApproach=Our, Params(M)=23.28, GFlops=96.04, FPS=31.92026.0677.9982.5866.66—
CPAApproach=Union, Params(M)=23.94, GFlops=217.91, FPS=13.22026.0677.6182.4365.89—
DENetApproach=Union, Params(M)=20.13, GFlops=65.33, FPS=45.02026.0677.582.165.74—
MGDIPApproach=Union, Params(M)=26.90, GFlops=116.45, FPS=8.12026.0677.3481.965.67—
RT-DETRApproach=Direct, Params(M)=20.09, GFlops=61.12, FPS=54.12026.0676.981.6664.85—
FPBackbone=ResNet-34, Resolution=300, FLOPs=68502019.0975.5———
FPBackbone=VGG-16, Resolution=300, FLOPs=299862019.0974.3———
BBGBackbone=VGG-16, Resolution=300, FLOPs=10622019.0968.5———
BDNBackbone=DenseNet-45, Resolution=512, FLOPs=19602019.0968.2———
BDNBackbone=DenseNet-37, Resolution=512, FLOPs=15302019.0966.4———
BBGBackbone=ResNet-34, Resolution=300, FLOPs=3622019.0962.8———
TBNBackbone=ResNet-34, Resolution=300, FLOPs=4642019.0959.5———
XNORBackbone=ResNet-34, Resolution=300, FLOPs=3622019.0955.1———
BARONTraining dataset=OV-LVIS, learnable prompt templates=true2024.06—7658.2—
DetProTraining dataset=OV-LVIS, Source=Reported from DetPro [4]2024.06—74.657.9—
LBPTraining dataset=OV-LVIS, learnable prompt templates=true2024.06—76.158.4—
SupervisedTraining dataset=Standard Supervised Labels2024.06—78.549—
ViLDTraining dataset=OV-LVIS, Source=Reported from DetPro [4]2024.06—73.957.9—