README.md

---
library_name: ultralytics
pipeline_tag: object-detection
license: cc-by-4.0
tags:
- yolo
- object-detection
- document-analysis
- pdf
- question-segmentation
- turkish
datasets:
- custom
base_model:
- yolo26n
---

# YOLO26N Question Segmentation

Smaller release checkpoint that slightly outperforms the YOLO26M release on both held-out test sets.

- Task: object detection
- Classes: `question`
- Input size: `1280`
- Base model: `yolo26n`

## Intended use

This model is designed to detect question regions in exam booklets, worksheets, and PDF page renders. It works best on dense educational pages with clear question boundaries.

## Release notes

This release checkpoint was trained for 50 epochs on the cleaned one-class question detection dataset.

Compared with the previously published HF YOLO11M baseline, this release improves held-out mAP50-95 by `+0.026` on the old test set and `+0.062` on the combined test set.

## Test results

### Old held-out test set

| Model | Size | Precision | Recall | mAP50 | mAP50-95 |
| --- | --- | --- | --- | --- | --- |
| [YOLO11M Question Segmentation v1](https://huggingface.co/erayyapagci/yolo11m-question-segmentation) | 116 MB / 20.1M params | 0.962 | 0.980 | 0.979 | 0.898 |
| [YOLO26M Question Segmentation v2](https://huggingface.co/erayyapagci/yolo26m-question-segmentation-v2) | 168 MB / 21.8M params | 0.990 | 0.982 | 0.988 | 0.920 |
| [YOLO26S Question Segmentation v2](https://huggingface.co/erayyapagci/yolo26s-question-segmentation-v2) | 20 MB / 9.9M params | 0.988 | 0.983 | 0.990 | 0.923 |
| [YOLO26N Question Segmentation v2](https://huggingface.co/erayyapagci/yolo26n-question-segmentation-v2) | 5.3 MB / 2.5M params | 0.988 | 0.983 | 0.989 | 0.924 |

### Combined held-out test set

| Model | Size | Precision | Recall | mAP50 | mAP50-95 |
| --- | --- | --- | --- | --- | --- |
| [YOLO11M Question Segmentation v1](https://huggingface.co/erayyapagci/yolo11m-question-segmentation) | 116 MB / 20.1M params | 0.962 | 0.973 | 0.984 | 0.900 |
| [YOLO26M Question Segmentation v2](https://huggingface.co/erayyapagci/yolo26m-question-segmentation-v2) | 168 MB / 21.8M params | 0.987 | 0.978 | 0.992 | 0.957 |
| [YOLO26S Question Segmentation v2](https://huggingface.co/erayyapagci/yolo26s-question-segmentation-v2) | 20 MB / 9.9M params | 0.988 | 0.991 | 0.993 | 0.963 |
| [YOLO26N Question Segmentation v2](https://huggingface.co/erayyapagci/yolo26n-question-segmentation-v2) | 5.3 MB / 2.5M params | 0.991 | 0.988 | 0.993 | 0.962 |

### Benchmark graphs

![Held-out mAP50-95](benchmark_map50_95.png)
![Held-out precision and recall](benchmark_precision_recall.png)

## Confidence guidance

From the confidence sweep used in this release:

- Best benchmark confidence for YOLO26S and YOLO26N is `conf=0.001` (highest mAP50-95 on both held-out test sets).
- For practical inference, start with `conf=0.25`; use `conf=0.20` if you want fewer missed borderline boxes.

## Qualitative examples

The newer YOLO26 checkpoints improve most clearly in two cases:

- Less false positives on pages with no questions.
- Better recovery of harder placements such as one-row questions and questions that occupy an entire column.

Example no-question page: GT + YOLO11M Question Segmentation v1 + YOLO26M Question Segmentation v2 + YOLO26S Question Segmentation v2 + YOLO26N Question Segmentation v2. The YOLO11M panel uses the earlier low-threshold prediction pass and shows two false positives; the YOLO26 checkpoints stay empty on the same page.

![2025 TYT no-question page comparison](comparison_examples/10__sourcepdf__yks_tyt_2025_kitapcik_d250__page_1.png)

## Usage

```python
from huggingface_hub import hf_hub_download
from ultralytics import YOLO

weights = hf_hub_download(repo_id="erayyapagci/yolo26n-question-segmentation-v2", filename="best.pt")
model = YOLO(weights)
results = model("page.png", imgsz=1280, conf=0.25)
```

Direct URL loading also works:

```python
from ultralytics import YOLO

model = YOLO("https://huggingface.co/erayyapagci/yolo26n-question-segmentation-v2/resolve/main/best.pt")
results = model("page.png", imgsz=1280, conf=0.25)
```

## Data note

Some of the training data comes from public Roboflow projects used in earlier versions of this question-segmentation pipeline:

1. **PDF Soru Cikarma** (tanimazsinu): [Link](https://universe.roboflow.com/tanimazsinu/pdf-soru-cikarma-wmwfu)
2. **WholeQuestionDetection** (Gazi University): [Link](https://universe.roboflow.com/gazi-university-gi7lx/wholequestiondetection)
3. **ExamBuddy** (ExamBuddy): [Link](https://universe.roboflow.com/exambuddy/exambuddy)
4. **Questions** (Terry Li): [Link](https://universe.roboflow.com/terry-li/questions-klidd)
5. **Question Parsing from Document** (Sefa): [Link](https://universe.roboflow.com/sefa-8tkmm/question-parsing-from-document)
6. **Question Dedector** (Nur Etinkaya): [Link](https://universe.roboflow.com/nur-etinkaya/question-dedector)
7. **Sorukes** (Sorualgilama): [Link](https://universe.roboflow.com/sorualgilama/sorukes-4akor)
8. **Question Detection** (Cognizen): [Link](https://universe.roboflow.com/cognizen/question-detection-aizzu)
9. **Questions2** (Fiver): [Link](https://universe.roboflow.com/fiver-1lcfr/questions2)
10. **Question-New** (Question): [Link](https://universe.roboflow.com/question/question-new)

## License

This release is published under `CC-BY-4.0`.