File size: 779 Bytes
f8800e2
 
 
 
 
 
 
 
8560caa
f8800e2
 
 
 
b639c98
f8800e2
b639c98
f8800e2
8560caa
b639c98
 
 
f8800e2
8560caa
b639c98
 
f8800e2
8560caa
f8800e2
 
 
8560caa
 
b639c98
f8800e2
8560caa
f8800e2
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
---
language: bn
license: mit
tags:
  - object-detection
  - ocr
  - bengali
  - yolov8
  - word-detection
metrics:
  - map
---

# Bengali OCR — Word-Level Detection (v3)

**Architecture:** YOLOv8n | **Task:** Word-level bounding box detection

## Results
| mAP@0.5 | Precision | Recall |
|---|---|---|
| 0.9223 | 0.9533 | 0.8722 |

## Training data
- ICDAR 2019 MLT Bengali (real word boxes)
- 6,000 synthetic printed pages (NID/form/paragraph style)

## Usage
```python
from ultralytics import YOLO
from huggingface_hub import hf_hub_download
path  = hf_hub_download("Sarjinkhan2003/bengali-ocr-detection", "bengali_det.pt")
model = YOLO(path)
results = model.predict("doc.jpg", conf=0.25)
for box in results[0].boxes:
    print(box.xyxy[0].tolist())  # one word per box
```