Home

Awesome

GearNet

AAAI 2022: GearNet: Stepwise Dual Learning for Weakly Supervised Domain Adaptation. (Pytorch implementation)

Requirements

Taining

All the commands are shown in scripts files.

Results

UNIF-20%A -> WA -> DW -> AW -> DD -> AD -> WAverage
Standard0.61970.63120.48610.92290.31640.80850.6308
Co-teaching0.61580.67910.56140.94370.54820.8450.6988
Jocor0.66530.74160.52370.97700.50280.90360.7190
DAN0.70310.71870.58090.95620.62990.91270.7502
DANN0.71220.75000.60930.96870.61570.92700.7638
TCL0.77730.81660.60750.98120.61000.93610.7881
G+Co-teaching0.73430.77910.60580.95000.60510.90360.7629
G+DANN0.75520.79160.61750.97290.61390.93610.7812
G+TCL0.81770.84370.61680.98750.62250.95180.8066
UNIF-40%A -> WA -> DW -> AW -> DD -> AD -> WAverage
Standard0.55850.58330.40980.81450.38700.73300.5810
Co-teaching0.59370.65830.49710.85000.33380.55980.5821
Jocor0.63020.66250.47230.95620.45240.83070.6673
DAN0.63800.67080.50710.89160.59510.89710.6999
DANN0.67310.70620.52230.92500.56960.88930.7142
TCL0.76560.75620.51060.94370.56000.88800.7373
G+Co-teaching0.69790.72500.55070.85830.34650.58330.6269
G+DANN0.74340.72290.53260.95000.57420.90490.7380
G+TCL0.79680.79160.49070.94580.53120.93220.7480
Flip-20%A -> WA -> DW -> AW -> DD -> AD -> WAverage
Standard0.60150.61870.44210.90620.40050.79290.6269
Co-teaching0.60540.61870.51910.86250.46090.70700.6289
Jocor0.64840.69370.48470.85410.34690.74210.6283
DAN0.66140.6770.55000.88120.58020.87100.7034
DANN0.66010.65410.53120.8770.57170.86450.6931
TCL0.75260.78330.57670.92290.60290.93220.76176
G+Co-teaching0.74730.67910.56710.87700.49890.7630.6887
G+DANN0.72650.73120.55390.88540.59510.89450.7311
G+TCL0.87760.82910.58730.93950.60720.94790.7981
Flip-40%A -> WA -> DW -> AW -> DD -> AD -> WAverage
Standard0.46870.48120.35470.72080.35040.63280.5014
Co-teaching0.50390.5250.36150.57290.29290.42570.4469
Jocor0.54290.6020.45950.71250.38920.69140.5662
DAN0.53120.54580.42180.71870.43530.66790.5534
DANN0.51430.50620.42320.69370.42220.67570.5392
TCL0.66010.61660.45950.76250.43350.64970.5969
G+Co-teaching0.57420.51040.38060.56660.31640.42830.4627
G+DANN0.54940.52080.42610.73540.43750.67310.5570
G+TCL0.69920.62700.46300.75410.42250.66010.6043

Citation

If you find this useful in your research, please consider citing:

@article{xie2022gearnet,
  title={GearNet: Stepwise Dual Learning for Weakly Supervised Domain Adaptation},
  author={Xie, Renchunzi and Wei, Hongxin and Feng, Lei and An, Bo},
  journal={AAAI},
  year={2022}
}

Contact

If you have any problems about our code, feel free to contact<br>

or describe your problem in Issues.