🧠 Submission to TGRS 2026：Curriculum-Aided Synergistic Self-Training for Source-Free Unsupervised Domain Adaptation of Remote Sensing Image Segmentation
🧠 Full source code will be released after the paper is accepted.
Source-free unsupervised domain adaptation (SFUDA) is a promising approach for remote sensing image (RSI) segmentation, as it transfers knowledge from a well-trained source network to an unlabeled target domain without requiring access to the source data. Most existing SFUDA methods rely on self-training to guide the transfer process. However, when applied to RSI segmentation, they still suffer from three unresolved challenges. First, due to large cross-domain discrepancies in RSIs, pseudo labels are often structuredly unreliable, making it difficult to distinguish trustworthy supervision from misleading predictions. Second, preserving representation diversity is difficult because reliability-oriented selection tends to discard informative hard regions and rare classes, resulting in insufficient target-domain coverage. Third, target samples exhibit significantly different adaptation difficulty, yet most methods optimize them with nearly uniform training strength, which often leads to unstable adaptation in the early stage. To address these challenges, we propose a Curriculum-aided Synergistic Self-training (CSS) framework, which performs selective pseudo-label learning in an easy-to-hard manner. Specifically, a Synergistic Self-Training (SST) strategy is proposed to achieve effective knowledge transfer by jointly modeling reliability and diversity. SST consists of two complementary paradigms: lazy learning and hungry learning. Lazy learning suppresses noise propagation by selecting pseudo labels with both high confidence and low uncertainty, while hungry learning improves target-domain representation coverage through global-wise and class-wise ordered label selection. Moreover, a Curriculum Information Propagation (CIP) strategy is proposed to stabilize the transfer process by progressively propagating domain-invariant information from easy samples to hard samples according to sample entropy. In this way, CSS establishes a unified adaptation mechanism that simultaneously addresses reliability, diversity, and stability. We further provide a theoretical analysis showing that CSS reduces the generalization error bound. Extensive experiments on satellite and aerial benchmarks demonstrate that CSS significantly outperforms state-of-the-art SFUDA methods and can be effectively extended to the black-box scenario.
- We propose the CSS framework for SFUDA semantic segmentation of RSIs, which explicitly addresses three unresolved challenges in remote sensing SFUDA: unreliable pseudo supervision, insufficient target representation diversity, and unstable adaptation across samples with different difficulty.
- We propose the SST strategy to achieve effective knowledge transfer. Unlike previous self-training methods that mainly emphasize either reliability filtering or broader pseudo-label utilization, SST unifies lazy learning for reliable feature learning and hungry learning for diverse feature learning in a complementary manner.
- We propose the CIP strategy to achieve stable knowledge transfer, which facilitates easy-to-hard information propagation by adaptively adjusting the supervision strength of lazy learning and hungry learning according to sample entropy.
- We provide theoretical proof that CSS narrows the domain gap by effectively reducing the generalization error bound. Extensive experimental results show that CSS outperforms state-of-the-art SFUDA methods on both satellite and aerial benchmarks. Moreover, CSS also exhibits competitive performance in the black-box scenario, where only the source network's predictions are available.
All datasets including ISPRS dataset and CASID dataset.
To train the source-only model:
CUDA_VISIBLE_DEVICES=0 python so_run.py
To train the adaptation model:
CUDA_VISIBLE_DEVICES=0 python run.py
python eval.py
| Domain | Method | SF | Surf | Bldg | Vegt | Tree | Car | Bkgd | mIoU | Domain | Surf | Bldg | Vegt | Tree | Car | Bkgd | mIoU |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| P2V | Source only | - | 28.4 | 54.0 | 21.9 | 50.8 | 28.0 | 2.0 | 30.9 | PRGB2V | 26.4 | 54.0 | 12.3 | 13.4 | 27.5 | 1.0 | 22.5 |
| AdaptSeg | ❌ | 54.4 | 63.1 | 29.0 | 52.7 | 6.4 | 4.6 | 35.0 | 51.3 | 60.7 | 12.8 | 51.6 | 10.3 | 3.0 | 31.6 | ||
| CBST | ❌ | 61.3 | 67.3 | 26.4 | 60.8 | 35.3 | 3.4 | 42.4 | 56.6 | 55.6 | 16.2 | 52.8 | 38.0 | 2.9 | 37.0 | ||
| ProDA | ❌ | 62.5 | 71.6 | 34.5 | 56.3 | 39.2 | 4.0 | 44.7 | 49.0 | 68.9 | 32.4 | 49.1 | 31.6 | 2.4 | 38.9 | ||
| DualGAN | ❌ | 49.4 | 62.3 | 38.9 | 57.7 | 34.3 | 29.7 | 45.4 | 46.2 | 65.4 | 27.9 | 55.8 | 40.3 | 3.9 | 40.0 | ||
| CCDA | ❌ | 67.7 | 76.8 | 47.0 | 55.0 | 44.9 | 20.7 | 52.0 | 64.5 | 76.9 | 38.4 | 52.8 | 43.4 | 12.4 | 48.1 | ||
| ProCA | ❌ | 64.6 | 76.5 | 41.8 | 64.0 | 30.8 | 8.3 | 47.7 | 54.2 | 64.2 | 17.7 | 65.2 | 36.5 | 4.1 | 40.3 | ||
| LD | ✅ | 54.1 | 50.9 | 34.7 | 44.1 | 19.8 | 5.5 | 33.8 | 49.4 | 43.7 | 23.0 | 32.4 | 15.0 | 3.5 | 28.0 | ||
| DTAC | ✅ | 52.9 | 52.5 | 32.7 | 42.5 | 20.0 | 5.8 | 33.4 | 48.5 | 42.3 | 19.0 | 32.2 | 15.2 | 3.8 | 26.8 | ||
| SND | ✅ | 54.5 | 53.3 | 34.4 | 43.3 | 21.5 | 6.9 | 34.7 | 49.6 | 44.0 | 20.0 | 31.0 | 17.4 | 3.8 | 27.8 | ||
| CROTS | ✅ | 53.8 | 51.1 | 34.3 | 43.9 | 20.3 | 5.6 | 33.9 | 49.7 | 44.9 | 21.2 | 32.1 | 17.6 | 3.9 | 28.1 | ||
| ATP | ✅ | 55.7 | 53.2 | 36.0 | 45.6 | 22.4 | 7.5 | 35.8 | 46.1 | 42.4 | 28.1 | 33.7 | 22.9 | 5.0 | 29.7 | ||
| SFDA-DE | ✅ | 54.4 | 53.0 | 34.3 | 43.9 | 21.6 | 6.9 | 34.7 | 50.9 | 43.9 | 22.9 | 32.3 | 19.6 | 4.2 | 29.0 | ||
| SFDA* | ✅ | 52.3 | 52.1 | 32.9 | 42.7 | 19.4 | 5.2 | 33.0 | 43.6 | 41.5 | 18.9 | 28.5 | 13.9 | 3.4 | 25.0 | ||
| HCL* | ✅ | 53.2 | 51.5 | 33.8 | 43.4 | 20.1 | 5.7 | 33.5 | 47.5 | 41.2 | 19.6 | 30.1 | 14.7 | 3.4 | 26.1 | ||
| CSS(Ours)* | ✅ | 56.0 | 48.3 | 33.5 | 44.5 | 21.4 | 6.7 | 35.1 | 51.1 | 46.3 | 22.2 | 32.9 | 18.2 | 3.9 | 29.1 | ||
| CSS(Ours) | ✅ | 55.3 | 51.1 | 36.0 | 42.4 | 25.1 | 6.1 | 36.0 | 50.4 | 43.1 | 24.2 | 35.2 | 24.7 | 3.9 | 30.2 | ||
| V2P | Source only | - | 51.3 | 47.3 | 33.9 | 19.2 | 46.0 | 13.1 | 35.1 | V2PRGB | 44.9 | 42.4 | 17.3 | 4.9 | 40.5 | 6.1 | 26.0 |
| AdaptSeg | ❌ | 49.6 | 48.0 | 34.4 | 22.6 | 41.0 | 8.4 | 34.0 | 37.7 | 54.3 | 15.1 | 30.7 | 42.3 | 6.1 | 31.0 | ||
| CBST | ❌ | 49.9 | 33.2 | 40.6 | 3.1 | 49.6 | 2.2 | 29.8 | 43.5 | 19.9 | 0.0 | 0.0 | 48.4 | 3.7 | 19.3 | ||
| ProDA | ❌ | 44.7 | 56.9 | 40.1 | 31.6 | 46.8 | 10.6 | 38.4 | 44.8 | 46.4 | 35.8 | 30.6 | 41.2 | 11.1 | 35.0 | ||
| DualGAN | ❌ | 51.0 | 53.4 | 36.5 | 35.0 | 48.5 | 11.5 | 39.3 | 46.0 | 59.0 | 41.7 | 25.8 | 39.7 | 13.6 | 37.6 | ||
| CCDA | ❌ | 64.4 | 66.4 | 47.2 | 37.6 | 59.4 | 12.3 | 47.9 | 57.7 | 65.4 | 29.8 | 35.9 | 57.0 | 13.3 | 43.2 | ||
| ProCA | ❌ | 57.4 | 40.6 | 45.6 | 12.3 | 63.1 | 4.8 | 37.3 | 40.1 | 27.1 | 5.8 | 2.7 | 58.0 | 7.1 | 23.5 | ||
| LD | ✅ | 44.3 | 49.0 | 30.7 | 37.0 | 50.8 | 8.9 | 36.8 | 37.4 | 38.6 | 15.5 | 29.9 | 42.6 | 2.9 | 27.8 | ||
| DTAC | ✅ | 39.0 | 45.6 | 30.6 | 35.0 | 45.2 | 7.8 | 33.9 | 30.9 | 36.1 | 14.9 | 28.1 | 37.3 | 2.2 | 24.9 | ||
| SND | ✅ | 43.8 | 48.5 | 31.6 | 36.3 | 50.4 | 8.2 | 36.5 | 33.3 | 39.4 | 22.9 | 35.4 | 44.5 | 2.2 | 29.6 | ||
| CROTS | ✅ | 43.2 | 47.0 | 34.5 | 37.2 | 47.6 | 7.8 | 36.2 | 33.1 | 34.5 | 16.1 | 28.1 | 38.4 | 1.8 | 25.4 | ||
| ATP | ✅ | 45.9 | 49.1 | 33.4 | 37.0 | 52.3 | 9.1 | 37.8 | 37.7 | 38.5 | 16.0 | 29.2 | 46.2 | 3.2 | 28.5 | ||
| SFDA-DE | ✅ | 39.6 | 47.9 | 30.8 | 36.9 | 46.2 | 9.6 | 35.2 | 36.0 | 36.1 | 18.6 | 30.9 | 34.0 | 5.2 | 26.6 | ||
| SFDA* | ✅ | 44.9 | 46.3 | 31.4 | 36.6 | 43.4 | 6.3 | 34.8 | 33.3 | 35.1 | 15.7 | 28.8 | 34.8 | 2.3 | 25.0 | ||
| HCL* | ✅ | 42.3 | 47.0 | 32.2 | 36.3 | 42.1 | 6.5 | 34.4 | 31.1 | 37.8 | 18.3 | 32.3 | 39.0 | 2.1 | 26.7 | ||
| CSS(Ours)* | ✅ | 46.0 | 46.5 | 35.4 | 30.2 | 49.1 | 7.0 | 35.7 | 35.4 | 38.0 | 16.2 | 30.2 | 41.4 | 4.4 | 27.6 | ||
| CSS(Ours) | ✅ | 50.3 | 48.9 | 31.9 | 35.2 | 53.4 | 7.7 | 37.9 | 43.8 | 38.8 | 14.9 | 29.9 | 50.3 | 2.5 | 30.0 |
| Domain | Method | SF | Bkgd | Bldg | Forest | Road | Water | mIoU | Domain | Bkgd | Bldg | Forest | Road | Water | mIoU |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| SubMs2TroMs | AdaptSeg | ❌ | 71.3 | 55.5 | 81.1 | 10.6 | 72.4 | 58.2 | SubMs2TroRf | 26.2 | 70.5 | 91.3 | 11.0 | 21.1 | 44.0 |
| CBST | ❌ | 72.1 | 54.0 | 82.0 | 16.5 | 68.7 | 58.7 | 24.6 | 71.6 | 92.2 | 25.4 | 20.4 | 46.8 | ||
| CLAN | ❌ | 72.3 | 56.3 | 81.2 | 13.6 | 67.0 | 58.1 | 26.1 | 71.6 | 93.2 | 14.7 | 34.8 | 48.1 | ||
| PyCDA | ❌ | 70.8 | 29.1 | 70.2 | 1.2 | 52.7 | 44.8 | 3.9 | 17.5 | 45.7 | 1.5 | 9.7 | 15.6 | ||
| FADA | ❌ | 72.5 | 50.1 | 80.6 | 7.9 | 46.5 | 51.6 | 20.0 | 65.7 | 90.5 | 12.5 | 28.0 | 43.3 | ||
| ASA | ❌ | 56.6 | 30.4 | 67.1 | 0.3 | 16.7 | 34.2 | 12.5 | 50.1 | 87.0 | 0.5 | 11.9 | 32.4 | ||
| LD | ✅ | 63.2 | 25.3 | 66.8 | 14.7 | 29.3 | 39.9 | 11.9 | 14.9 | 72.1 | 5.5 | 9.8 | 22.8 | ||
| DTAC | ✅ | 62.2 | 24.3 | 66.7 | 14.7 | 25.3 | 38.6 | 10.7 | 16.2 | 72.5 | 4.4 | 9.9 | 22.7 | ||
| SND | ✅ | 62.6 | 23.5 | 66.4 | 15.7 | 26.8 | 39.0 | 11.3 | 18.2 | 72.5 | 4.9 | 8.4 | 23.1 | ||
| CROTS | ✅ | 62.1 | 23.1 | 65.9 | 15.3 | 22.7 | 37.8 | 12.6 | 16.7 | 75.7 | 5.4 | 11.0 | 24.3 | ||
| ATP | ✅ | 63.2 | 28.2 | 69.6 | 15.9 | 26.5 | 40.7 | 11.8 | 15.6 | 72.6 | 5.4 | 9.7 | 23.0 | ||
| SFDA-DE | ✅ | 62.8 | 26.0 | 67.9 | 14.2 | 25.8 | 39.3 | 11.0 | 13.3 | 68.8 | 4.6 | 8.7 | 21.2 | ||
| SFDA* | ✅ | 62.3 | 22.6 | 64.6 | 15.9 | 24.6 | 38.0 | 10.6 | 16.2 | 70.7 | 4.9 | 8.5 | 21.9 | ||
| HCL* | ✅ | 62.2 | 21.1 | 63.0 | 14.5 | 25.9 | 37.3 | 11.9 | 14.9 | 72.0 | 5.8 | 9.7 | 22.9 | ||
| CSS(Ours)* | ✅ | 62.6 | 24.8 | 66.7 | 15.5 | 26.2 | 39.2 | 12.1 | 16.4 | 74.4 | 4.8 | 9.9 | 23.6 | ||
| CSS(Ours) | ✅ | 63.3 | 30.3 | 70.9 | 15.4 | 27.9 | 41.6 | 12.3 | 17.7 | 75.9 | 5.9 | 11.0 | 24.6 | ||
| TemMs2TroMs | AdaptSeg | ❌ | 59.5 | 53.1 | 75.7 | 2.9 | 36.6 | 45.6 | TemMs2TroRf | 7.5 | 66.2 | 83.3 | 5.7 | 14.5 | 35.4 |
| CBST | ❌ | 70.1 | 53.2 | 81.0 | 2.3 | 39.5 | 49.2 | 6.5 | 66.0 | 58.1 | 18.7 | 5.6 | 31.0 | ||
| CLAN | ❌ | 64.0 | 56.2 | 81.9 | 3.5 | 42.6 | 49.7 | 8.6 | 64.3 | 81.5 | 8.9 | 28.0 | 38.2 | ||
| PyCDA | ❌ | 59.5 | 20.3 | 47.8 | 0.6 | 19.1 | 29.5 | 4.8 | 31.4 | 46.8 | 0.8 | 4.9 | 17.8 | ||
| FADA | ❌ | 62.5 | 48.8 | 75.4 | 4.5 | 29.8 | 44.2 | 7.1 | 48.0 | 67.4 | 8.1 | 11.4 | 28.4 | ||
| ASA | ❌ | 50.9 | 19.8 | 70.3 | 0.5 | 51.7 | 38.6 | 7.6 | 29.8 | 83.0 | 1.8 | 0.1 | 24.5 | ||
| LD | ✅ | 53.3 | 46.7 | 71.1 | 14.2 | 23.5 | 41.7 | 6.3 | 23.8 | 69.6 | 4.0 | 7.5 | 22.3 | ||
| DTAC | ✅ | 51.9 | 45.2 | 71.0 | 12.9 | 21.4 | 39.8 | 7.0 | 32.5 | 74.3 | 5.2 | 6.9 | 25.1 | ||
| SND | ✅ | 53.0 | 44.4 | 72.7 | 13.4 | 26.6 | 42.0 | 6.9 | 29.8 | 70.6 | 4.6 | 6.9 | 23.9 | ||
| CROTS | ✅ | 52.6 | 46.1 | 71.7 | 13.6 | 21.6 | 41.1 | 6.9 | 30.0 | 72.6 | 4.5 | 6.2 | 24.6 | ||
| ATP | ✅ | 53.8 | 46.9 | 72.3 | 13.0 | 23.9 | 41.6 | 6.8 | 33.9 | 71.9 | 4.7 | 6.4 | 24.7 | ||
| SFDA-DE | ✅ | 54.4 | 45.8 | 72.1 | 12.7 | 22.4 | 41.5 | 6.6 | 30.1 | 70.8 | 4.5 | 6.2 | 23.7 | ||
| SFDA* | ✅ | 51.1 | 46.0 | 71.2 | 13.8 | 19.1 | 40.3 | 6.4 | 34.1 | 70.9 | 5.0 | 7.2 | 24.8 | ||
| HCL* | ✅ | 51.1 | 45.1 | 71.5 | 12.4 | 20.5 | 40.1 | 7.0 | 26.6 | 77.0 | 4.9 | 7.5 | 24.6 | ||
| CSS(Ours)* | ✅ | 52.7 | 47.0 | 71.9 | 13.9 | 21.4 | 41.4 | 6.5 | 34.8 | 72.5 | 5.7 | 7.4 | 25.4 | ||
| CSS(Ours) | ✅ | 53.5 | 46.4 | 73.1 | 14.6 | 26.2 | 42.8 | 7.4 | 35.4 | 78.3 | 5.5 | 8.1 | 26.9 | ||
| TroRf2TroMs | AdaptSeg | ❌ | 65.7 | 56.1 | 79.2 | 7.0 | 74.8 | 56.5 | TroMs2TroRf | 17.9 | 58.2 | 85.0 | 18.2 | 29.1 | 41.7 |
| CBST | ❌ | 66.3 | 61.2 | 78.9 | 6.8 | 68.9 | 56.4 | 17.1 | 56.9 | 85.1 | 22.1 | 22.0 | 40.6 | ||
| CLAN | ❌ | 66.3 | 56.7 | 78.7 | 5.4 | 76.6 | 56.8 | 18.2 | 69.9 | 85.4 | 20.3 | 29.0 | 44.6 | ||
| PyCDA | ❌ | 64.0 | 16.6 | 46.6 | 0.4 | 57.4 | 37.0 | 4.6 | 18.6 | 53.6 | 1.5 | 17.2 | 19.1 | ||
| FADA | ❌ | 52.0 | 42.9 | 73.6 | 6.5 | 69.2 | 48.8 | 19.0 | 65.2 | 87.4 | 19.9 | 22.3 | 42.8 | ||
| ASA | ❌ | 52.5 | 28.2 | 66.1 | 5.0 | 33.5 | 36.1 | 9.1 | 50.4 | 81.5 | 2.6 | 23.8 | 33.5 | ||
| LD | ✅ | 45.3 | 42.0 | 72.9 | 9.8 | 35.8 | 41.3 | 11.4 | 16.1 | 71.2 | 5.6 | 20.3 | 24.9 | ||
| DTAC | ✅ | 46.6 | 42.2 | 72.6 | 9.2 | 38.4 | 41.8 | 10.9 | 15.4 | 70.1 | 4.6 | 20.6 | 24.5 | ||
| SND | ✅ | 45.3 | 40.0 | 73.0 | 9.0 | 37.3 | 41.0 | 10.5 | 13.0 | 71.4 | 4.2 | 18.5 | 24.2 | ||
| CROTS | ✅ | 43.2 | 41.7 | 72.5 | 9.2 | 36.3 | 40.6 | 11.0 | 16.1 | 72.0 | 4.6 | 19.4 | 25.0 | ||
| ATP | ✅ | 47.9 | 42.2 | 73.3 | 8.9 | 39.8 | 42.4 | 11.5 | 15.7 | 74.0 | 4.9 | 21.8 | 26.4 | ||
| SFDA-DE | ✅ | 46.0 | 42.8 | 73.7 | 9.2 | 34.1 | 42.0 | 10.9 | 15.9 | 73.7 | 5.4 | 21.2 | 26.0 | ||
| SFDA* | ✅ | 45.3 | 40.3 | 72.7 | 9.0 | 31.7 | 39.8 | 10.3 | 13.0 | 71.9 | 4.6 | 18.1 | 23.0 | ||
| HCL* | ✅ | 45.3 | 42.1 | 72.8 | 9.2 | 35.2 | 41.1 | 10.9 | 15.4 | 70.0 | 4.5 | 20.5 | 23.3 | ||
| CSS(Ours)* | ✅ | 47.1 | 42.7 | 73.3 | 10.3 | 36.1 | 41.9 | 10.9 | 16.6 | 74.8 | 5.3 | 21.2 | 26.7 | ||
| CSS(Ours) | ✅ | 49.0 | 43.0 | 72.9 | 9.9 | 42.5 | 43.5 | 12.1 | 23.1 | 78.1 | 5.4 | 23.4 | 28.4 | ||
| SubMs2TemMs | AdaptSeg | ❌ | 44.4 | 60.2 | 45.3 | 21.9 | 6.4 | 35.7 | TemMs2SubMs | 40.3 | 76.6 | 74.8 | 22.5 | 51.7 | 53.2 |
| CBST | ❌ | 37.7 | 54.0 | 23.8 | 27.7 | 4.3 | 29.5 | 40.9 | 75.7 | 66.3 | 25.8 | 29.7 | 47.7 | ||
| CLAN | ❌ | 41.1 | 59.7 | 36.0 | 22.6 | 7.2 | 33.3 | 45.5 | 76.9 | 77.4 | 23.5 | 39.7 | 52.6 | ||
| PyCDA | ❌ | 37.3 | 32.3 | 22.8 | 7.2 | 0.3 | 20.0 | 13.3 | 57.5 | 44.0 | 4.4 | 17.9 | 27.4 | ||
| FADA | ❌ | 48.4 | 60.1 | 59.8 | 19.5 | 5.0 | 38.6 | 40.0 | 75.6 | 70.9 | 21.9 | 49.1 | 51.5 | ||
| ASA | ❌ | 34.2 | 28.2 | 60.0 | 9.9 | 4.9 | 27.4 | 24.2 | 53.0 | 62.6 | 66.0 | 13.7 | 30.8 | ||
| LD | ✅ | 38.3 | 30.0 | 37.3 | 7.5 | 2.0 | 23.0 | 32.7 | 65.4 | 62.2 | 14.6 | 21.4 | 39.3 | ||
| DTAC | ✅ | 32.3 | 29.0 | 41.2 | 4.7 | 2.2 | 21.9 | 31.8 | 60.5 | 61.8 | 13.4 | 15.1 | 36.5 | ||
| SND | ✅ | 36.0 | 31.0 | 40.3 | 7.1 | 2.5 | 23.4 | 32.6 | 60.9 | 63.1 | 13.7 | 19.6 | 38.0 | ||
| CROTS | ✅ | 36.8 | 28.5 | 38.2 | 7.2 | 2.6 | 22.7 | 33.1 | 61.0 | 62.3 | 13.4 | 16.2 | 37.2 | ||
| ATP | ✅ | 38.4 | 23.4 | 44.6 | 11.7 | 3.5 | 24.5 | 33.5 | 65.2 | 62.6 | 14.1 | 21.9 | 39.5 | ||
| SFDA-DE | ✅ | 37.1 | 32.3 | 40.3 | 8.7 | 3.1 | 24.3 | 32.1 | 63.9 | 61.8 | 14.6 | 19.4 | 38.4 | ||
| SFDA* | ✅ | 37.2 | 25.8 | 36.2 | 1.2 | 2.6 | 20.6 | 31.4 | 61.7 | 63.4 | 14.6 | 17.5 | 37.7 | ||
| HCL* | ✅ | 38.9 | 29.7 | 33.1 | 4.1 | 3.1 | 21.8 | 32.3 | 62.5 | 62.2 | 14.0 | 19.6 | 38.1 | ||
| CSS(Ours)* | ✅ | 42.0 | 29.2 | 33.3 | 11.1 | 2.6 | 23.6 | 32.7 | 66.0 | 62.1 | 14.2 | 21.7 | 39.4 | ||
| CSS(Ours) | ✅ | 40.0 | 28.7 | 46.8 | 10.1 | 1.6 | 25.4 | 33.8 | 65.8 | 66.3 | 13.9 | 24.5 | 40.9 | ||
| TroMs2TemMs | AdaptSeg | ❌ | 42.3 | 44.1 | 29.7 | 25.1 | 0.7 | 28.4 | TroMs2SubMs | 50.5 | 75.7 | 73.2 | 27.7 | 37.0 | 52.8 |
| CBST | ❌ | 45.2 | 39.1 | 27.2 | 28.0 | 0.3 | 28.0 | 46.6 | 70.7 | 64.7 | 32.5 | 35.6 | 50.0 | ||
| CLAN | ❌ | 41.8 | 50.1 | 34.9 | 26.6 | 1.7 | 31.0 | 50.2 | 77.9 | 73.8 | 29.8 | 39.1 | 54.1 | ||
| PyCDA | ❌ | 37.7 | 32.0 | 19.6 | 1.9 | 0.2 | 18.3 | 37.1 | 73.0 | 44.9 | 4.5 | 9.3 | 33.8 | ||
| FADA | ❌ | 45.7 | 47.1 | 41.3 | 19.3 | 0.6 | 30.8 | 54.1 | 73.8 | 80.6 | 16.6 | 46.7 | 54.4 | ||
| ASA | ❌ | 40.0 | 31.7 | 36.9 | 1.8 | 0.2 | 22.1 | 41.2 | 65.6 | 69.4 | 1.1 | 16.7 | 38.8 | ||
| LD | ✅ | 36.0 | 20.4 | 39.8 | 3.3 | 2.2 | 20.4 | 42.3 | 55.0 | 60.7 | 6.4 | 29.2 | 38.7 | ||
| DTAC | ✅ | 37.5 | 17.2 | 30.8 | 3.5 | 2.4 | 18.5 | 42.0 | 49.0 | 56.7 | 5.8 | 24.9 | 35.7 | ||
| SND | ✅ | 40.6 | 22.8 | 31.7 | 8.2 | 2.4 | 21.1 | 42.2 | 54.5 | 59.4 | 5.8 | 23.7 | 37.1 | ||
| CROTS | ✅ | 39.5 | 22.8 | 31.7 | 8.4 | 2.3 | 20.9 | 42.2 | 49.2 | 57.4 | 5.7 | 24.9 | 35.9 | ||
| ATP | ✅ | 39.4 | 20.2 | 44.9 | 7.9 | 2.7 | 22.8 | 42.2 | 55.1 | 62.6 | 6.1 | 29.3 | 39.1 | ||
| SFDA-DE | ✅ | 41.0 | 24.2 | 34.9 | 9.9 | 2.3 | 22.5 | 41.4 | 51.0 | 60.6 | 5.6 | 29.4 | 37.6 | ||
| SFDA* | ✅ | 43.7 | 16.5 | 31.6 | 5.8 | 2.6 | 20.1 | 42.1 | 50.8 | 56.4 | 6.0 | 23.7 | 35.8 | ||
| HCL* | ✅ | 38.8 | 22.8 | 37.2 | 2.2 | 2.5 | 20.7 | 42.1 | 51.6 | 59.3 | 6.1 | 27.1 | 37.2 | ||
| CSS(Ours)* | ✅ | 40.2 | 23.0 | 34.8 | 9.4 | 2.4 | 21.9 | 43.2 | 52.5 | 60.2 | 6.0 | 28.0 | 38.0 | ||
| CSS(Ours) | ✅ | 41.1 | 22.2 | 44.4 | 8.1 | 2.2 | 23.6 | 43.6 | 55.8 | 63.7 | 6.2 | 35.8 | 41.0 | ||
| TroRf2TemMs | AdaptSeg | ❌ | 38.9 | 59.2 | 65.4 | 24.2 | 4.2 | 38.2 | TroRf2SubMs | 28.9 | 75.3 | 72.2 | 22.8 | 60.1 | 51.9 |
| CBST | ❌ | 36.1 | 57.5 | 59.2 | 27.9 | 1.3 | 36.4 | 42.9 | 77.5 | 72.5 | 32.0 | 27.8 | 50.5 | ||
| CLAN | ❌ | 37.6 | 60.3 | 51.9 | 25.5 | 1.3 | 35.8 | 27.0 | 76.8 | 70.8 | 26.6 | 48.2 | 49.9 | ||
| PyCDA | ❌ | 30.6 | 20.3 | 25.7 | 1.2 | 0.2 | 15.6 | 35.2 | 69.5 | 42.3 | 4.0 | 8.0 | 31.8 | ||
| FADA | ❌ | 40.1 | 57.1 | 64.9 | 19.3 | 1.4 | 36.6 | 24.5 | 74.2 | 70.9 | 18.4 | 38.9 | 45.4 | ||
| ASA | ❌ | 37.4 | 18.7 | 53.0 | 46.0 | 4.0 | 27.9 | 35.5 | 63.7 | 71.4 | 0.0 | 1.6 | 34.4 | ||
| LD | ✅ | 33.3 | 28.7 | 39.0 | 5.6 | 1.7 | 21.7 | 35.4 | 67.5 | 69.3 | 10.6 | 30.3 | 42.6 | ||
| DTAC | ✅ | 36.2 | 29.6 | 33.1 | 5.4 | 1.5 | 21.2 | 32.9 | 65.9 | 68.6 | 9.7 | 30.1 | 41.4 | ||
| SND | ✅ | 30.6 | 34.6 | 43.2 | 8.2 | 1.9 | 23.7 | 34.3 | 63.5 | 68.4 | 9.8 | 29.9 | 41.2 | ||
| CROTS | ✅ | 34.0 | 31.8 | 41.4 | 8.4 | 1.8 | 23.5 | 35.6 | 68.6 | 70.2 | 10.2 | 31.1 | 43.1 | ||
| ATP | ✅ | 23.7 | 38.6 | 54.6 | 5.8 | 2.6 | 25.1 | 33.6 | 66.4 | 69.3 | 9.9 | 32.1 | 42.2 | ||
| SFDA-DE | ✅ | 35.5 | 32.9 | 37.3 | 9.9 | 1.8 | 23.5 | 33.7 | 68.4 | 69.6 | 10.0 | 31.8 | 42.7 | ||
| SFDA* | ✅ | 23.1 | 22.5 | 43.1 | 1.7 | 1.7 | 18.6 | 34.2 | 65.1 | 68.2 | 10.2 | 26.4 | 40.8 | ||
| HCL* | ✅ | 34.3 | 26.9 | 37.9 | 3.5 | 1.8 | 20.9 | 34.2 | 63.9 | 67.8 | 9.3 | 24.9 | 40.0 | ||
| CSS(Ours)* | ✅ | 33.3 | 35.3 | 43.3 | 9.2 | 1.5 | 24.5 | 35.2 | 65.6 | 68.9 | 9.6 | 30.3 | 41.9 | ||
| CSS(Ours) | ✅ | 27.2 | 41.3 | 52.6 | 8.4 | 1.2 | 26.1 | 35.8 | 67.4 | 70.1 | 10.0 | 35.8 | 43.8 |
Our implementation is mainly based on following repositories. Thanks for their authors.
If you encounter any problems or bugs, please don't hesitate to contact me at yiweifang@hhu.edu.cn.