| --- |
| license: cc-by-nc-sa-4.0 |
| --- |
| |
| # SeC: Advancing Complex Video Object Segmentation via Progressive Concept Construction |
|
|
| [\[📂 GitHub\]](https://github.com/OpenIXCLab/SeC) |
| [\[📦 Model\]](https://huggingface.co/OpenIXCLab/SeC-4B) |
| [\[🌐 Homepage\]](https://rookiexiong7.github.io/projects/SeC/) |
| [\[📄 Paper\]](https://arxiv.org/abs/2507.xxxxx) |
|
|
| ## Highlights |
|
|
| - 🔥We introduce **Segment Concept (SeC)**, a **concept-driven** segmentation framework for **video object segmentation** that integrates **Large Vision-Language Models (LVLMs)** for robust, object-centric representations. |
| - 🔥SeC dynamically balances **semantic reasoning** with **feature matching**, adaptively adjusting computational efforts based on **scene complexity** for optimal segmentation performance. |
| - 🔥We propose the **Semantic Complex Scenarios Video Object Segmentation (SeCVOS)** benchmark, designed to evaluate segmentation in challenging scenarios. |
|
|
| ## SeCVOS Benchmark |
|
|
| We propose the Semantic Complex Scenarios Video Object Segmentation (SeCVOS) benchmark, specifically designed to assess a model’s ability to perform high-level semantic reasoning across complex visual narratives. SeCVOS contains 160 carefully curated multi-shot videos characterized by: 1) Highly discontinuous frame sequences, 2) Frequent reappearance of objects across disparate scenes, and 3) Abrupt shot transitions and dynamic camera motion. |
|
|
| | Benchmark | #Videos | Avg. Duration (s) | Disapp. Rate | Avg. #Scene | |
| | :------------------------------ | :------: | :---------------: | :----------: | :-----------: | |
| | DAVIS | 90 | 2.87 | 16.1% | 1.06 | |
| | YTVOS | 507 | 4.51 | 13.0% | 1.03 | |
| | MOSE | 311 | 8.68* | 41.5% | 1.06 | |
| | SA-V | 155 | 17.24 | 25.5% | 1.09 | |
| | LVOS | 140 | 78.36 | 7.8% | 1.47 | |
| | **SeCVOS (ours)** | 160 | 29.36 | 30.2% | **4.26** | |
|
|
| ## License |
| Our annotations are licensed under a [CC BY-NC-SA 4.0 License](https://creativecommons.org/licenses/by-nc-sa/4.0/). They are available strictly for non-commercial research. |
|
|
| We uphold the rights of individuals and copyright holders. If you are featured in any of our video annotations or hold copyright to a video and wish to have its annotation removed from our dataset, please reach out to us. Send an email to zhangzhixiong@pjlab.org.cn with the subject line beginning with SeCVOS, or raise an issue with the same title format. We commit to reviewing your request promptly and taking suitable action. |
|
|
