Specification Parameters
Basic Information
- Model
- doubao-seed-2-1-pro-260915
- Key Parameters
- bbox 归一化坐标 1000×1000, 3D bbox 输出 9 参数, point 计数, polygon 文字定位
资料下载
Comparison with Similar Products
| Specifications | 本品 · doubao-seed-2-1-pro-260915 | kimi-k3.1-pro | 火山方舟模型 API | qwen-audio-3.1-asr-flash | doubao-seed-2.1-turbo |
|---|---|---|---|---|---|
| Main controller | — | — | — | — | — |
| NPU | — | — | — | — | — |
| Memory / VRAM | — | — | — | — | — |
| Reference unit price | Request a Quote | Request a Quote | Request a Quote | Request a Quote | Request a Quote |
| Inventory | — | — | — | — | — |
Product introduction
Visual Grounding is a visual understanding capability introduced in the Volcano Ark documentation: based on natural language task requirements, it locates the corresponding target in an image and returns coordinates. The model must possess both visual understanding and natural language understanding capabilities; compared with object detection, it adds natural language input, making the interaction more flexible. Coordinates are returned as ratio coordinates normalized to 1000×1000 and must be converted according to the actual image dimensions. It is suitable for scenarios such as object localization, multi-target detection, text position extraction, counting, and 3D box output; before use, you need to activate the Ark platform and obtain an API Key.