Hugging Face's logo

Datasets:
syCen
/
CameraBench

Modalities:
Image
Text
Formats:
json
Size:
1K - 10K
ArXiv:
Libraries:
Datasets
pandas
License:
Dataset card Data Studio Files
xet
Community
2
syCen commited on
Commit
35a9d04
·
verified ·
1 Parent(s): 54eeab0

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +1 -1
README.md CHANGED
@@ -11,7 +11,7 @@ license: mit
11
  [![](https://img.shields.io/badge/%F0%9F%8F%A0%20_Homepage-4285F4?color=4285F4&logoColor=white)](https://linzhiqiu.github.io/papers/camerabench/)
12
  [![](https://img.shields.io/badge/%F0%9F%A4%97%20_CameraBench_testset-FF9B00?color=FF9B00&logoColor=white)](https://huggingface.co/datasets/syCen/CameraBench)
13
 
14
- ![Demo GIF](./images/sfm_vs_vlm.jpg)
15
  > **SfMs and VLMs performance on CameraBench**: Generative VLMs (evaluated with [VQAScore](https://linzhiqiu.github.io/papers/vqascore/)) trail classical SfM/SLAM in pure geometry, yet they outperform discriminative VLMs that rely on CLIPScore/ITMScore and—even better—capture scene‑aware semantic cues missed by SfM
16
  > After simple supervised fine‑tuning (SFT) on ≈1,400 extra annotated clips, our 7B Qwen2.5‑VL doubles its AP, outperforming the current best MegaSAM.
17
 
 
11
  [![](https://img.shields.io/badge/%F0%9F%8F%A0%20_Homepage-4285F4?color=4285F4&logoColor=white)](https://linzhiqiu.github.io/papers/camerabench/)
12
  [![](https://img.shields.io/badge/%F0%9F%A4%97%20_CameraBench_testset-FF9B00?color=FF9B00&logoColor=white)](https://huggingface.co/datasets/syCen/CameraBench)
13
 
14
+ ![Demo GIF](https://raw.githubusercontent.com/sy77777en/CameraBench/main/images/sfm_vs_vlm.jpg)
15
  > **SfMs and VLMs performance on CameraBench**: Generative VLMs (evaluated with [VQAScore](https://linzhiqiu.github.io/papers/vqascore/)) trail classical SfM/SLAM in pure geometry, yet they outperform discriminative VLMs that rely on CLIPScore/ITMScore and—even better—capture scene‑aware semantic cues missed by SfM
16
  > After simple supervised fine‑tuning (SFT) on ≈1,400 extra annotated clips, our 7B Qwen2.5‑VL doubles its AP, outperforming the current best MegaSAM.
17