Add task category and improve metadata

README.md

@@ -1,16 +1,18 @@
 ---
-license: apache-2.0
 language:
 - en
 - cn
+license: apache-2.0
+size_categories:
+- 1K<n<10K
+pretty_name: OK-VOS
+task_categories:
+- image-segmentation
 tags:
 - multimodal
 - search
 - video-object-segmentation
 - reasoning
-pretty_name: OK-VOS
-size_categories:
-- 1K<n<10K
 ---
 
 # OK-VOS: A Video Object Segmentation Benchmark Requiring Outside Knowledge
@@ -39,7 +41,15 @@ data/OK_VOS/
 
 ## ⚙️ How to Use
 
-Please refer to our repo: https://github.com/iSEE-Laboratory/Seg-ReSearch
+To prepare the dataset for training or evaluation using the scripts from the [official repository](https://github.com/iSEE-Laboratory/Seg-ReSearch), you can run the following preprocessing commands:
+
+```bash
+# Training Set Preprocessing
+python examples/data_preprocess/okvos.py --split train --num_frames 6 --max_size 448 --min_size 448 --model_type Qwen3VL
+
+# Test set Preprocessing
+python examples/data_preprocess/okvos.py --split test --num_frames 6 --max_size 448 --min_size 448 --model_type Qwen3VL
+```
 
 ## 📜 Citation
 If you find this dataset useful, please cite our paper:
@@ -51,4 +61,4 @@ author={Tianming Liang and Qirui Du and Jian-Fang Hu and Haichao Jiang and Ziche
 journal={arXiv preprint arXiv:2602.04454},
 year={2026}
 }
-```
+```