| <div align="center"> | |
| <h1>Context as Memory: Scene-Consistent Interactive Long Video Generation with Memory Retrieval</h1> | |
| <h1>SIGGRAPH Asia 2025</h1> | |
| <p> | |
| <a href="https://context-as-memory.github.io/">[Project page]</a> | |
| <a href="https://arxiv.org/pdf/2506.03141">[ArXiv]</a> | |
| <a href="https://huggingface.co/datasets/KwaiVGI/Context-as-Memory-Dataset">[Dataset]</a> | |
| </p> | |
| </div> | |
| # File Structure | |
| To prepare the dataset for use, merge the parts into a single zip file using the following command: | |
| ```bash | |
| cat Context-as-Memory-Dataset_* > Context-as-Memory-Dataset.zip | |
| ``` | |
| After extracting `Context-as-Memory-Dataset.zip`, the dataset will be organized as follows: | |
| ``` | |
| Context-as-Memory-Dataset | |
| ├── frames | |
| │ ├── AncientTempleEnv_0 | |
| │ │ ├── 0000.png | |
| │ │ ├── 0001.png | |
| │ │ ├── 0002.png | |
| │ │ └── ... | |
| │ ├── AncientTempleEnv_1 | |
| │ │ ├── 0000.png | |
| │ │ ├── 0001.png | |
| │ │ ├── 0002.png | |
| │ │ └── ... | |
| │ └── ... | |
| │ | |
| ├── jsons | |
| │ ├── AncientTempleEnv_0.json | |
| │ ├── AncientTempleEnv_1.json | |
| │ └── ... | |
| │ | |
| ├── overlap_labels | |
| │ ├── AncientTempleEnv_0 | |
| │ │ ├── 0.json | |
| │ │ ├── 1.json | |
| │ │ ├── 2.json | |
| │ │ └── ... | |
| │ ├── AncientTempleEnv_1 | |
| │ │ ├── 0.json | |
| │ │ ├── 1.json | |
| │ │ ├── 2.json | |
| │ │ └── ... | |
| │ └── ... | |
| │ | |
| └── captions.txt | |
| ``` | |
| # Explanation of Dataset Parts | |
| - **`frames/`**: 100 subdirectories, each containing 7,601 video frame images. | |
| - **`jsons/`**: 100 JSON files, each storing the camera pose (position + rotation) of every frame in the corresponding long video. | |
| - **`overlap_labels/`**: 100 subdirectories, each containing 7,601 JSON files, where each file records the indices of overlapping frames corresponding to that frame. | |
| - **`captions.txt`**: Captions annotated for a segment of a long video, from a given starting frame to an ending frame. | |
| - We also provide a simple code file, `tools.py`, which can convert (x, y, z, yaw, pitch) into RT, and can also select a specific frame as the reference frame to align the RT of other frames to its coordinate system. | |