Object Detection on DOTA HBB v2.0 (val)
47.14mAPSLIP-RS*
Evaluation results
| Method | Links | |
|---|---|---|
| SLIP-RS*Backbone=ConvNeXT-L, Training data=RS-Attribute-15M, Evaluation Protocol=Fine-tune2026.05 | 47.14 | |
| SLIP-RSBackbone=ConvNeXT-T, Training data=RS-O + RS-C, Evaluation Protocol=Fine-tune2026.05 | 45.1 | |
| OpenRSDBackbone=RTMDet-L, Training data=ORSD+, Evaluation Protocol=Fine-tune2026.05 | 44.57 | |
| SLIP-RS*Backbone=ConvNeXT-T, Training data=RS-Attribute-15M, Evaluation Protocol=Fine-tune2026.05 | 44.43 | |
| SLIP-RSBackbone=ConvNeXT-T, Training data=RS-C, Evaluation Protocol=Fine-tune2026.05 | 42.56 | |
| SLIP-RSBackbone=ConvNeXT-T, Training data=RS-O, Evaluation Protocol=Fine-tune2026.05 | 41.24 | |
| RTMDetBackbone=RTMDet-L (Lyu et al., 2022), Training data=DOTA2.0 & DIOR, Evaluation Protocol=Fine-tune2026.05 | 39.43 | |
| LAE-DINOBackbone=Swin-T, Training data=LAE-1M, Evaluation Protocol=Fine-tune2026.05 | 39.38 | |
| Faster-RCNNBackbone=ConvNeXT-T (Liu et al., 2022), Training data=DOTA2.0 & DIOR, Evaluation Protocol=Fine-tune2026.05 | 39.21 | |
| DINOBackbone=Swin-T (Liu et al., 2021), Training data=DOTA2.0 & DIOR, Evaluation Protocol=Fine-tune2026.05 | 38.32 | |
| SLIP-RSBackbone=ConvNeXT-T, Training data=RS-C, Evaluation Protocol=Zero-shot2026.05 | 35.63 |