Update README.md

README.md

@@ -29,21 +29,28 @@ The dataset includes x2 and x4 scaling pairs in `.npy` format, suitable for trai
     - `eval_hr_patch/`, `eval_lr_patch/` (100 files each)
   - Croissant metadata: `sampled_data/x2/croissant.json`
 
-## Loading Sample
+
+## Dataset Viewer Configuration
+To preview the sample data in Dataset Viewer, use the following Python script to load `.npy` files:
+
 ```python
 from datasets import Dataset
 import numpy as np
 import glob
 
-# Load paired data
-train_hr_files = glob.glob("sampled_data/x2/train_hr_patch/*.npy")
+train_hr_files = sorted(glob.glob("sampled_data/x2/train_hr_patch/*.npy"))
 train_lr_files = [f.replace("hr_patch", "lr_patch") for f in train_hr_files]
-eval_hr_files = glob.glob("sampled_data/x2/eval_hr_patch/*.npy")
+eval_hr_files = sorted(glob.glob("sampled_data/x2/eval_hr_patch/*.npy"))
 eval_lr_files = [f.replace("hr_patch", "lr_patch") for f in eval_hr_files]
 data_dict = {
-    "hr_data": [np.load(f) for f in train_hr_files + eval_hr_files],
-    "lr_data": [np.load(f) for f in train_lr_files + eval_lr_files],
+    "hr_attn_map": [np.load(f, allow_pickle=True).item()["attn_map"] for f in train_hr_files + eval_hr_files],
+    "hr_mask": [np.load(f, allow_pickle=True).item()["mask"] for f in train_hr_files + eval_hr_files],
+    "hr_image": [np.load(f, allow_pickle=True).item()["image"] for f in train_hr_files + eval_hr_files],
+    "hr_coord": [np.load(f, allow_pickle=True).item()["coord"] for f in train_hr_files + eval_hr_files],
+    "lr_attn_map": [np.load(f, allow_pickle=True).item()["attn_map"] for f in train_lr_files + eval_lr_files],
+    "lr_mask": [np.load(f, allow_pickle=True).item()["mask"] for f in train_lr_files + eval_lr_files],
+    "lr_image": [np.load(f, allow_pickle=True).item()["image"] for f in train_lr_files + eval_lr_files],
+    "lr_coord": [np.load(f, allow_pickle=True).item()["coord"] for f in train_lr_files + eval_lr_files],
     "split": ["train"] * len(train_hr_files) + ["eval"] * len(eval_hr_files)
 }
 dataset = Dataset.from_dict(data_dict)
-print(dataset[0]["hr_data"].shape)  # Example access