Table 5. Results on Cityscapes, using models trained on SIM10K (from SIM10K to Cityscapes) (%).
Method | img | ins | CTX | L | mimg | mins | mcons | Car AP |
---|---|---|---|---|---|---|---|---|
Faster R-CNN(baseline) | 34.3 | |||||||
DA Model [9] | ✓ | ✓ | 39.4 | |||||
SW-DA [10] | ✓ | ✓ | ✓ | 40.1 | ||||
SW-DA(γ = 3) [10] | ✓ | ✓ | 42.3 | |||||
SC-DA(Type3) [11] | 43.0 | |||||||
SLNO-S1(ours) | ✓ | ✓ | ✓ | 41.3 | ||||
SLNO-S2(ours) | ✓ | ✓ | ✓ | 41.6 | ||||
SLNO-S3(ours) | ✓ | ✓ | ✓ | ✓ | 42.4 | |||
SLNO-S4(ours) | ✓ | ✓ | ✓ | ✓ | ✓ | 42.7 | ||
SLNO(λ = 0.15)(ours) | ✓ | ✓ | ✓ | ✓ | ✓ | 43.0 |