Upload folder using huggingface_hub

.gitattributes

@@ -58,3 +58,8 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 # Video files - compressed
 *.mp4 filter=lfs diff=lfs merge=lfs -text
 *.webm filter=lfs diff=lfs merge=lfs -text
+Scene-Decoupled-Video-Dataset.part_aa filter=lfs diff=lfs merge=lfs -text
+Scene-Decoupled-Video-Dataset.part_ab filter=lfs diff=lfs merge=lfs -text
+Scene-Decoupled-Video-Dataset.part_ac filter=lfs diff=lfs merge=lfs -text
+Scene-Decoupled-Video-Dataset.part_ad filter=lfs diff=lfs merge=lfs -text
+Scene-Decoupled-Video-Dataset.part_ae filter=lfs diff=lfs merge=lfs -text

Scene-Decoupled-Video-Dataset.part_aa

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ec9a3cff8cc09751598a415a9539a515a990fb4b110e27976a29872bb92ca65e
+size 42949672960

Scene-Decoupled-Video-Dataset.part_ab

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3d24e5b5998d655bb9d287f95912de3a1f7872631c8fc397431ef955f23f0ebc
+size 42949672960

Scene-Decoupled-Video-Dataset.part_ac

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:76a0853116a7e9887360ea0cfafa7ec5ed186a4149cf7b80f1437217dd621b2c
+size 42949672960

Scene-Decoupled-Video-Dataset.part_ad

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:031acc67c0e964ac1bfd7c77dd58d2bdf2694db8eeb593daad6024a151ff8055
+size 42949672960

Scene-Decoupled-Video-Dataset.part_ae

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a74f873d5dcea0c07331663b36ba6fdfe741e7d673f8e6e5f1c8fef773d4fc71
+size 35741068577

extract_scene_from_panorama.py

@@ -0,0 +1,79 @@
+import os
+import numpy as np
+from PIL import Image
+from equilib import equi2pers
+
+# --- Configuration ---
+# Path to the input equirectangular (panorama) image
+INPUT_IMAGE_PATH = "panorama/scene1_3x3_loc1_scene_AncientTempleEnv/scene1_3x3_loc1_scene_AncientTempleEnv_pano.jpeg"
+# Directory where the output perspective frames will be saved
+OUTPUT_DIR = "panorama/scene1_3x3_loc1_scene_AncientTempleEnv/extracted"
+
+# Camera parameters
+NUM_FRAMES = 20
+HEIGHT = 384
+WIDTH = 672
+FOV_X = 90.0  
+
+def generate_perspective_frames():
+    """
+    Extracts a sequence of perspective frames from an equirectangular image
+    by rotating the yaw angle.
+    """
+    # 1. Prepare Output Directory
+    if not os.path.exists(OUTPUT_DIR):
+        os.makedirs(OUTPUT_DIR)
+        print(f"Created directory: {OUTPUT_DIR}")
+
+    # 2. Load and Preprocess Image
+    if not os.path.exists(INPUT_IMAGE_PATH):
+        print(f"Error: Input file not found at {INPUT_IMAGE_PATH}")
+        return
+
+    equi_img = Image.open(INPUT_IMAGE_PATH).convert("RGB")
+    equi_img_np = np.asarray(equi_img)
+    
+    # equilib requires (C, H, W) format
+    equi_img_chw = np.transpose(equi_img_np, (2, 0, 1))
+
+    print(f"Successfully loaded: {INPUT_IMAGE_PATH}")
+    print(f"Generating {NUM_FRAMES} frames...")
+
+    # 3. Calculate Yaw Angles
+    yaw_angles = np.linspace(0, 2 * np.pi, num=NUM_FRAMES, endpoint=False)
+
+    # 4. Processing Loop
+    for i, yaw in enumerate(yaw_angles):
+        # Define rotation (in radians)
+        # pitch=0: horizontal view; roll=0: no tilt
+        rots = {
+            'roll': 0.0,
+            'pitch': 0.0,
+            'yaw': yaw,
+        }
+
+        # Transform equirectangular to perspective
+        pers_img_chw = equi2pers(
+            equi=equi_img_chw,
+            rots=rots,
+            height=HEIGHT,
+            width=WIDTH,
+            fov_x=FOV_X,
+            mode="bilinear",
+        )
+
+        # Convert back to (H, W, C) and uint8 for saving
+        pers_img_hwc = np.transpose(pers_img_chw, (1, 2, 0)).astype(np.uint8)
+        output_image = Image.fromarray(pers_img_hwc)
+
+        # Save the frame
+        filename = f"frame_{str(i).zfill(3)}.jpg"
+        save_path = os.path.join(OUTPUT_DIR, filename)
+        output_image.save(save_path)
+        
+        print(f"Saved: {filename} | Yaw: {np.rad2deg(yaw):.2f}°")
+
+    print("\nProcessing complete. All frames generated successfully.")
+
+if __name__ == "__main__":
+    generate_perspective_frames()

readme.md

@@ -0,0 +1,87 @@
+# Scene-Decoupled Video Dataset
+
+TL;DR: The Scene-Decoupled Video Dataset, introduced in CineScene, is a large-scale synthetic dataset for **video generation with decoupled scene**, which encompasses diverse scenes, subjects, and camera movements. This dataset contains camera trajectories, equirectangular panorama (scene image), and videos with/without dynamic subject. The data is organized into "With Human" (whuman) and "Without Human" (wohuman) categories, while panoramas are scene-decoupled and shared across both.
+
+## 1. Directory Tree
+
+```text
+.
+├── camera/                           # Camera trajectories and metadata
+│   ├── whuman/                       # Sequences containing human characters
+│   │   └── <scene_id>/               # e.g., scene1_3x3_loc1_scene_AncientTempleEnv/
+│   │       └── <scene_id>_cam.json   # Camera parameters
+│   └── wohuman/                      # Sequences with environment only
+│       └── <scene_id>/
+│           └── <scene_id>_cam.json
+│
+├── panorama/                         # Scene-decoupled environment maps
+│   └── <scene_id>/                   # Shared between whuman and wohuman
+│       └── <scene_id>_pano.jpeg      # 360° Equirectangular panoramic image
+│
+└── video/                            # Rendered video sequences (MP4)
+    ├── whuman/                       # Videos with human characters
+    │   └── <scene_id>/
+    │       ├── <scene_id>_01_24mm.mp4 # Sub-sequences (01, 02, etc.)
+    │       ├── <scene_id>_02_24mm.mp4
+    │       └── ...
+    └── wohuman/                      # Videos without human characters
+        └── <scene_id>/
+            ├── <scene_id>_01_24mm.mp4
+            ├── ...
+
+```
+
+
+
+## 2. Dataset Statistics
+
+*   **Total Scale**: 46,816 videos.
+*   **Scenes**: 6,688 scenes (comprising 3,344 *whuman* and 3,344 *wohuman* scenes) across 35 high-quality 3D environments.
+*   **Trajectories**: 46,816 camera paths (7 distinct camera trajectories per scene).
+*   **Panorama**: 360° Equirectangular images for every scene, providing a complete background reference for scene conditioning.
+
+| Property | Value |
+| :--- | :--- |
+| **Video Resolution** | 672 x 384 |
+| **Frame Count** | 81 frames per video |
+| **Frame Rate** | 15 FPS |
+| **View Change Range** | Up to 75° |
+| **Decoupled Scene** | 360° Equirectangular (Panorama) |
+| **Panorama Resolution** | 2048 x 1024 |
+
+
+## 3. Dataset Construction
+
+We follow the asset collection pipeline established by **RecamMaster**, but introduce three significant enhancements to support more complex generative tasks:
+
+1.  **Decoupled Scenes**: We provide static 360° panoramic images (Equirectangular) for every scene. This allows for explicit background conditioning and facilitates novel view synthesis from any angle.
+2.  **Extended Camera Range**: Our dataset covers significantly larger view changes (approx. **75°**) compared to the 5–60° range provided in previous datasets [1].
+3.  **Paired Subject/Background Data**: Every scene includes both "with-subject" (*whuman*) and "background-only" (*wohuman*) video sequences. This paired data is ideal for training models on subject-background decoupling, motion transfer, and cinematic composition.
+
+## 4. useful script
+
+- download
+
+```bash
+sudo apt-get install git-lfs
+git lfs install
+git clone https://huggingface.co/datasets/KwaiVGI/Scene-Decoupled-Video-Dataset
+cat Scene-Decoupled-Video-Dataset.part* > Scene-Decoupled-Video-Dataset.tar.gz
+tar -xvf Scene-Decoupled-Video-Dataset.tar.gz
+```
+
+- camera visualization
+
+  To visualize the camera, please refer to [here.](https://huggingface.co/datasets/KlingTeam/MultiCamVideo-Dataset/blob/main/vis_cam.py)
+
+- Perspective Projection
+To extract perspective frames from the panoramic images:
+
+  ```
+  python extract_scene_from_panorama.py
+  ```
+
+  
+
+References
+[1] Bai J, Xia M, Fu X, et al. Recammaster: Camera-controlled generative rendering from a single video[J]. arXiv preprint arXiv:2503.11647, 2025.