Update README.md
Browse files
README.md
CHANGED
|
@@ -14,10 +14,13 @@ tags:
|
|
| 14 |
|
| 15 |
Compared to [MolDet](https://huggingface.co/UniParser/MolDet), **MolDetv2** leverages more manually annotated training data, with further optimizations specifically for reducing molecular false detections and improving bounding box regression, achieving stronger performance with a smaller model.
|
| 16 |
|
|
|
|
| 17 |
## [MolDet-General] General molecule structure detection models
|
| 18 |
|
| 19 |
YOLO11-n weights trained on more than 100k human annotated image crops & synthesis molecule images.
|
| 20 |
|
|
|
|
|
|
|
| 21 |
* 640x640 input resolution
|
| 22 |
* support handwritten molecules detection
|
| 23 |
* **multiscale input** (inputs can be single/multiple molecular cutouts, reaction or table cutouts, or single-page PDF images)
|
|
@@ -31,10 +34,14 @@ model.predict("path/to/image.png", save=True, imgsz=640, conf=0.5)
|
|
| 31 |
```
|
| 32 |
For further usage instructions, please refer to the [official Ultralytics documentation](https://docs.ultralytics.com/modes/predict/).
|
| 33 |
|
|
|
|
|
|
|
| 34 |
## [MolDet-Doc] PDF molecule structure detection models
|
| 35 |
|
| 36 |
YOLO11-n weights trained on more than 60k human annotated PDF pages (patents, papers, and books) and 10k synthesis PDF pages with molecule images.
|
| 37 |
|
|
|
|
|
|
|
| 38 |
* 960x960 input resolution
|
| 39 |
* prefer **single page PDF image** input
|
| 40 |
* better in small molecule detection
|
|
@@ -55,6 +62,7 @@ for i, p in enumerate(pdf):
|
|
| 55 |
```
|
| 56 |
For further usage instructions, please refer to the [official Ultralytics documentation](https://docs.ultralytics.com/modes/predict/).
|
| 57 |
|
|
|
|
| 58 |
## 📊 BenchMark Results
|
| 59 |
|
| 60 |
Please refer to [MolDet-Bench](https://huggingface.co/datasets/UniParser/MolDet-Bench)
|
|
@@ -69,5 +77,5 @@ Please refer to [MolDet-Bench](https://huggingface.co/datasets/UniParser/MolDet-
|
|
| 69 |
If you use this model in your work, please cite:
|
| 70 |
|
| 71 |
```
|
| 72 |
-
|
| 73 |
```
|
|
|
|
| 14 |
|
| 15 |
Compared to [MolDet](https://huggingface.co/UniParser/MolDet), **MolDetv2** leverages more manually annotated training data, with further optimizations specifically for reducing molecular false detections and improving bounding box regression, achieving stronger performance with a smaller model.
|
| 16 |
|
| 17 |
+
|
| 18 |
## [MolDet-General] General molecule structure detection models
|
| 19 |
|
| 20 |
YOLO11-n weights trained on more than 100k human annotated image crops & synthesis molecule images.
|
| 21 |
|
| 22 |
+
|
| 23 |
+
|
| 24 |
* 640x640 input resolution
|
| 25 |
* support handwritten molecules detection
|
| 26 |
* **multiscale input** (inputs can be single/multiple molecular cutouts, reaction or table cutouts, or single-page PDF images)
|
|
|
|
| 34 |
```
|
| 35 |
For further usage instructions, please refer to the [official Ultralytics documentation](https://docs.ultralytics.com/modes/predict/).
|
| 36 |
|
| 37 |
+
|
| 38 |
+
|
| 39 |
## [MolDet-Doc] PDF molecule structure detection models
|
| 40 |
|
| 41 |
YOLO11-n weights trained on more than 60k human annotated PDF pages (patents, papers, and books) and 10k synthesis PDF pages with molecule images.
|
| 42 |
|
| 43 |
+
|
| 44 |
+
|
| 45 |
* 960x960 input resolution
|
| 46 |
* prefer **single page PDF image** input
|
| 47 |
* better in small molecule detection
|
|
|
|
| 62 |
```
|
| 63 |
For further usage instructions, please refer to the [official Ultralytics documentation](https://docs.ultralytics.com/modes/predict/).
|
| 64 |
|
| 65 |
+
|
| 66 |
## 📊 BenchMark Results
|
| 67 |
|
| 68 |
Please refer to [MolDet-Bench](https://huggingface.co/datasets/UniParser/MolDet-Bench)
|
|
|
|
| 77 |
If you use this model in your work, please cite:
|
| 78 |
|
| 79 |
```
|
| 80 |
+
Comming soon!
|
| 81 |
```
|