| license: apache-2.0 | |
| task_categories: | |
| - summarization | |
| language: | |
| - en | |
| # 📚 Clip-CC Dataset | |
| **Clip-CC** is a high-quality dataset of 200 short educational videos (1 minute 30 seconds each) paired with human-written summaries. It’s designed for tasks like video summarization, video-language alignment, and multimodal learning. | |
| --- | |
| ## 🧾 Dataset Summary | |
| - 🎞️ 200 videos: Named `1.mp4` to `200.mp4` | |
| - 📝 Each video has a concise human-written summary | |
| - 📂 File references stored in a JSONL file (`metadata.jsonl`) with paths to each video | |
| --- | |
| ## 💡 Use Cases | |
| - Video summarization and caption generation | |
| - Vision-Language alignment | |
| - Video QA and downstream educational AI tasks | |
| - Fine-tuning multimodal models (e.g., Flamingo, Video-BERT, LLaVA) | |
| --- | |
| ## 📦 Dataset Structure | |
| Each entry has: | |
| | Field | Description | | |
| |------------|------------------------------------------| | |
| | `id` | Unique ID, e.g., `001` | | |
| | `file_name`| Relative path to video file | | |
| | `summary` | Human-written summary of video content | | |
| ### 🔍 Example | |
| ```json | |
| { | |
| "id": "001", | |
| "file_name": "clips/1.mp4", | |
| "summary": "An introduction to gravitational force using animations." | |
| } | |