Instructions to use moon-g/DocLayout-YOLO-DocStructBench-onnx with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- YOLOv10
How to use moon-g/DocLayout-YOLO-DocStructBench-onnx with YOLOv10:
from ultralytics import YOLOvv10 model = YOLOvv10.from_pretrained("moon-g/DocLayout-YOLO-DocStructBench-onnx") source = 'http://images.cocodataset.org/val2017/000000039769.jpg' model.predict(source=source, save=True) - Notebooks
- Google Colab
- Kaggle
Add model card
Browse files
README.md
ADDED
|
@@ -0,0 +1,55 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: apache-2.0
|
| 3 |
+
base_model:
|
| 4 |
+
- juliozhao/DocLayout-YOLO-DocStructBench
|
| 5 |
+
language:
|
| 6 |
+
- en
|
| 7 |
+
pipeline_tag: object-detection
|
| 8 |
+
tags:
|
| 9 |
+
- layout-analysis
|
| 10 |
+
- document-understanding
|
| 11 |
+
- yolov10
|
| 12 |
+
- onnx
|
| 13 |
+
---
|
| 14 |
+
|
| 15 |
+
# DocLayout-YOLO-DocStructBench ONNX(moon-reader 镜像)
|
| 16 |
+
|
| 17 |
+
moon-reader(小月亮阅读器)使用的 DocLayout-YOLO 版面识别模型 ONNX 镜像。
|
| 18 |
+
|
| 19 |
+
## 来源
|
| 20 |
+
|
| 21 |
+
由 [wybxc/DocLayout-YOLO-DocStructBench-onnx](https://huggingface.co/wybxc/DocLayout-YOLO-DocStructBench-onnx) 转存,原模型转换自 [juliozhao/DocLayout-YOLO-DocStructBench](https://huggingface.co/juliozhao/DocLayout-YOLO-DocStructBench)(Apache-2.0)。
|
| 22 |
+
|
| 23 |
+
本仓库为 moon-reader 应用内模型下载的稳定镜像源,避免上游仓库变动影响应用。
|
| 24 |
+
|
| 25 |
+
## 文件
|
| 26 |
+
|
| 27 |
+
| 文件 | 大小 | SHA256 |
|
| 28 |
+
|---|---|---|
|
| 29 |
+
| `doclayout_yolo_docstructbench_imgsz1024.onnx` | 75,324,598 B | `fece9af02f618b603ff7921ccec6861d13e7e1f9830e091dfb7e8ad9311e5b21` |
|
| 30 |
+
|
| 31 |
+
## 模型规格
|
| 32 |
+
|
| 33 |
+
- 架构:Ultralytics YOLOv10m
|
| 34 |
+
- 任务:版面检测(10 类)
|
| 35 |
+
- 输入:`[1, 3, 1024, 1024]` f32,RGB→BGR 转换 + letterbox(pad 114)+ 归一化 `/255`
|
| 36 |
+
- 输出:`[1, 300, 6]`(x1, y1, x2, y2, score, class_id),端到端 NMS
|
| 37 |
+
|
| 38 |
+
### 类别顺序(class id)
|
| 39 |
+
|
| 40 |
+
```
|
| 41 |
+
0: title
|
| 42 |
+
1: plain text
|
| 43 |
+
2: abandon
|
| 44 |
+
3: figure
|
| 45 |
+
4: figure_caption
|
| 46 |
+
5: table
|
| 47 |
+
6: table_caption
|
| 48 |
+
7: table_footnote
|
| 49 |
+
8: isolate_formula
|
| 50 |
+
9: formula_caption
|
| 51 |
+
```
|
| 52 |
+
|
| 53 |
+
## 使用
|
| 54 |
+
|
| 55 |
+
应用内自动下载(国内源优先,SHA256 校验)。推理用 tract(纯 Rust)或 ONNX Runtime。
|