Create README.md
Browse files
README.md
ADDED
|
@@ -0,0 +1,22 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: gpl-3.0
|
| 3 |
+
tags:
|
| 4 |
+
- cross-modal-retrieval
|
| 5 |
+
- point-cloud
|
| 6 |
+
base_model:
|
| 7 |
+
- facebook/dinov2-large
|
| 8 |
+
- google/tipsv2-l14
|
| 9 |
+
- Pointcept/Utonia
|
| 10 |
+
---
|
| 11 |
+
|
| 12 |
+
# CrossWeave: Cross-Modal Scene and Instance Retrieval from Sparse 2D-3D Alignment
|
| 13 |
+
|
| 14 |
+
**NeurIPS 2026** · Paper (coming soon) · [Project page](https://crossweave3d.github.io) · [Code](https://github.com/crossweave3d/crossweave)
|
| 15 |
+
|
| 16 |
+
| File | Image encoder | Scene R@1 ScanNet | Scene R@1 3RScan | Instance Top-1 ScanNet | Instance Top-1 3RScan |
|
| 17 |
+
|---|---|---|---|---|---|
|
| 18 |
+
| `crossweave_dinov2_vitl14.ckpt` | DINOv2 ViT-L/14 | 60.58 | 47.13 | 77.66 | 73.61 |
|
| 19 |
+
| `crossweave_tipsv2_l14.ckpt` | TIPSv2 ViT-L/14 | 59.29 | 43.95 | 75.22 | 73.28 |
|
| 20 |
+
|
| 21 |
+
Image → point cloud, val split. Checkpoints contain only the trained adapters and SALAD. Installation, data
|
| 22 |
+
preparation and evaluation are described in the [code repository](https://github.com/crossweave3d/crossweave).
|