add dataset info

README.md

@@ -1,15 +1,66 @@
-# IMCPT-SparseGM-generator
+# IMCPT-SparseGM
+
+IMCPT-SparseGM dataset is a new visual graph matching benchmark addressing partial matching and graphs with larger sizes, based on the novel stereo benchmark [Image Matching Challenge PhotoTourism  (IMC-PT)  2020](https://www.cs.ubc.ca/research/image-matching-challenge/2020/). This dataset is released in CVPR 2023 paper *Deep Learning of Partial Graph Matching via Differentiable Top-K*.
+
+A comparison of existing vision graph matching datasets is presented:
+
+#### Comparison of Existing Vision Graph Matching Datasets
+
+| **dataset name**        | **# images** | **# classes** | **avg # nodes** | **# universe** | **partial rate** | **data type** |
+| ----------------------- | ------------ | ------------- | --------------- | -------------- | ---------------- | ------------- |
+| **CMU house/hotel**     | 212          | 2             | 30              | 30             | 0.0%             | gray-scale    |
+| **Willow ObjectClass**  | 404          | 5             | 10              | 10             | 0.0%             | RGB           |
+| **CUB2011**             | 11788        | 200           | 12.0            | 15             | 20.0%            | RGB           |
+| **Pascal VOC Keypoint** | 8702         | 20            | 9.07            | 6 to 23        | 28.5%            | RGB           |
+| **IMC-PT-SparseGM-50**  | 25061        | 16            | 21.36           | 50             | 57.3%            | RGB           |
+
+A visualization of 3D point cloud labels provided by **the original IMC-PT (blue)** and our selected anchor points for graph matching in **IMC-PT-SparseGM (red)**:
+
+![reichstag-3D-selected](./dataset_imgs/reichstag-3D-selected.png)
+
+A visualization of graph matching labels from **IMC-PT-SparseGM**:
+
+![reichstag-visual](./dataset_imgs/reichstag-visual.png)
+
+
+
+### IMCPT-SparseGM-generator
+
 This generator creates IMCPT-SparseGM based on Image_Matching_Challange_Data.
 
-Note that you should install colmap and download Image_Matching_Challange_Data before you create IMCPT-SparseGM by just running 
-
+Note that you should install colmap and download [Image_Matching_Challange_Data](https://www.cs.ubc.ca/~kmyi/imw2020/data.html) before you create IMCPT-SparseGM by just running 
     python dataset_generator.py
 
 Arguments are the following:
 
-    --root             'source dataset directory'                             default='/mnt/nas/dataset_share/Image_Matching_Challange_Data'
+    --root             'source dataset directory'                             default='Image_Matching_Challange_Data'
     --out_dir          'output dataset directory'                             default='picture'
     --pt_num           'universal point number to be selected'                default=50
     --min_exist_num    'min num of img an anchor exists in'                   default=10
     --dis_rate         'min distance rate when selecting points'              default=1.0
     --exist_dis_rate   'min distance rate when judging anchors\' existence'   default=0.75
+
+We provide the download links of IMC-PT-SparseGM-50, i.e., IMC-PT-SparseGM with 50 anchor points from [google drive](https://drive.google.com/file/d/1Po9pRMWXTqKK2ABPpVmkcsOq-6K_2v-B/view?usp=sharing) or [baidu drive (code: 0576)](https://pan.baidu.com/s/1hlJdIFp4rkiz1Y-gztyHIw).
+
+You can also generate IMC-PT-SparseGM annotations by your demands (such as setting ``pt_num`` to 100), using IMCPT-SparseGM-generator.
+
+
+
+Please cite the following papers if you use IMC-PT-SparseGM dataset:
+
+```
+@article{JinIJCV21,
+  title={Image Matching across Wide Baselines: From Paper to Practice},
+  author={Jin, Yuhe and Mishkin, Dmytro and Mishchuk, Anastasiia and Matas, Jiri and Fua, Pascal and Yi, Kwang Moo and Trulls, Eduard},
+  journal={International Journal of Computer Vision},
+  pages={517--547},
+  year={2021}
+}
+
+@unpublished{WangCVPR23,
+  title={Deep Learning of Partial Graph Matching via Differentiable Top-K},
+  author={Runzhong Wang*, Ziao Guo*, Shaofei Jiang, Xiaokang Yang, Junchi Yan},
+  booktitle={CVPR},,
+  year={2023}
+}
+```

dataset_generator.py

@@ -36,7 +36,7 @@ def get_image(idx):
 if __name__=='__main__':
     parser = argparse.ArgumentParser()
     parser.add_argument('--root', dest='root', help='source dataset directory',
-                        default='/mnt/nas/dataset_share/Image_Matching_Challange_Data', 
+                        default='Image_Matching_Challange_Data', 
                         type=str)
     parser.add_argument('--out_dir', dest='out_dir', help='output dataset directory',
                         default='picture',