---
license: apache-2.0
language:
  - en
  - de
tags:
  - object-detection
  - yolo
  - yolov8
  - ultralytics
  - invoice
  - document-layout
  - document-understanding
  - ocr-prep
  - invoice-extraction
library_name: ultralytics
pipeline_tag: object-detection
base_model: ultralytics/yolov8x
datasets:
  - AvoCahDoe/invoice-annotated-bbox
metrics:
  - map_50
  - map
  - precision
  - recall
model-index:
  - name: best
    results:
      - task:
          type: object-detection
        dataset:
          name: invoice-yolo-annotated
          type: invoice-layout
        metrics:
          - type: map_50
            value: 0.8757
          - type: map
            value: 0.6224
          - type: precision
            value: 0.9144
          - type: recall
            value: 0.8735
---

# Invoice Layout Detection — YOLOv8x

Fine-tuned **YOLOv8x** for **8-class invoice document layout** region detection (metadata, vendor/customer blocks, table, line items, summary, payment, column headers).

Part of the Hub collection **[Invoice Layout Extraction](https://huggingface.co/collections/AvoCahDoe/invoice-layout-extraction-6a39cd2a061a508c8049d825)**.

| Property | Value |
|----------|-------|
| **Architecture** | YOLOv8x (68.2M params, 257.8 GFLOPs) |
| **Base weights** | `ultralytics/yolov8x` |
| **Input size** | 1024 px |
| **Classes** | 8 layout regions |
| **Training epochs** | 100 (best @ epoch 100) |
| **Optimizer** | AdamW |
| **Dataset** | [AvoCahDoe/invoice-annotated-bbox](https://huggingface.co/datasets/AvoCahDoe/invoice-annotated-bbox) |
| **Demo Space** | [AvoCahDoe/invoice-layout-yolov8n-demo](https://huggingface.co/spaces/AvoCahDoe/invoice-layout-yolov8n-demo) |
| **Train / val / test** | 372 / 14 / 7 pages |

## Classes

| ID | Name |
|----|------|
| 0 | `invoice_metadata` |
| 1 | `vendor_block` |
| 2 | `customer_block` |
| 3 | `table_block` |
| 4 | `line_item` |
| 5 | `summary_block` |
| 6 | `payment_block` |
| 7 | `Column` |

## Metrics

### Test split (held-out, 7 images)

| Metric | Value |
|--------|-------|
| **mAP50** | 0.8757 |
| **mAP50-95** | 0.6224 |
| **Precision** | 0.9144 |
| **Recall** | 0.8735 |

### Validation (best epoch 100)

| Metric | Best | Final (epoch 100) |
|--------|------|-----------------------------------------------|
| mAP50 | 0.8767 | 0.8767 |
| mAP50-95 | 0.6016 | 0.6016 |
| Precision | 0.9213 | 0.9213 |
| Recall | 0.7415 | 0.7415 |

Training time: ~85.4 min on RTX 4070 Laptop GPU.

## Training configuration

| Parameter | Value |
|-----------|-------|
| Batch | 2 |
| Image size | 1024 |
| Patience | 25 |
| LR (cosine) | 0.01 → 0.01 |
| Mosaic | 1.0 |
| Mixup | 0.15 |
| Copy-paste | 0.1 |
| Horizontal flip | 0.0 |

## Training results

### Learning curves

![Training results](assets/results.png)

### Precision–Recall

![PR curve](assets/BoxPR_curve.png)

### Confusion matrices

| Raw | Normalized |
|-----|------------|
| ![Confusion matrix](assets/confusion_matrix.png) | ![Normalized confusion matrix](assets/confusion_matrix_normalized.png) |

### Validation predictions

| Ground truth | Model predictions |
|--------------|-------------------|
| ![Val labels](assets/val_batch0_labels.jpg) | ![Val predictions](assets/val_batch0_pred.jpg) |

### Training batches (augmented)

| Batch 0 | Batch 1 | Batch 2 |
|---------|---------|---------|
| ![Train batch 0](assets/train_batch0.jpg) | ![Train batch 1](assets/train_batch1.jpg) | ![Train batch 2](assets/train_batch2.jpg) |

### Label distribution

![Label distribution](assets/labels.jpg)

## Usage

### Ultralytics (recommended)

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

weights = hf_hub_download("AvoCahDoe/invoice-layout-yolov8x", "weights/best.pt")
model = YOLO(weights)
results = model.predict("invoice_page.png", imgsz=1024, conf=0.25)
results[0].show()
```

### Load from Hub by repo id

```python
from ultralytics import YOLO

model = YOLO("hf://AvoCahDoe/invoice-layout-yolov8x/weights/best.pt")
results = model.predict("invoice_page.png")
```

### Training reproduction

```bash
python scripts/train_comparison.py --models yolov8x
```

## Repository layout

```
weights/best.pt          # Best checkpoint (use this)
weights/last.pt          # Last epoch checkpoint
config/                  # Training configuration
metrics/                 # Per-epoch and summary metrics
assets/                  # Plots and visualizations
```

## Model comparison (test split)

All six architectures fine-tuned on the same invoice layout dataset.

| Rank | Model | test mAP50 | test mAP50-95 | Precision | Recall | Hub |
|------|-------|------------|---------------|-----------|--------|-----|
| 1 | `yolov8n` | 0.9600 | 0.7167 | 0.9502 | 0.9592 | [YOLOv8n](https://huggingface.co/AvoCahDoe/invoice-layout-yolov8n) |
| 2 | `yolov8s` | 0.9006 | 0.6739 | 0.9419 | 0.8677 | [YOLOv8s](https://huggingface.co/AvoCahDoe/invoice-layout-yolov8s) |
| 3 | `yolov8x` **← this model** | 0.8757 | 0.6224 | 0.9144 | 0.8735 | [YOLOv8x](https://huggingface.co/AvoCahDoe/invoice-layout-yolov8x) |
| 4 | `yolo11x` | 0.8738 | 0.6127 | 0.9484 | 0.8530 | [YOLO11x](https://huggingface.co/AvoCahDoe/invoice-layout-yolo11x) |
| 5 | `yolo11m` | 0.8418 | 0.4926 | 0.9033 | 0.8466 | [YOLO11m](https://huggingface.co/AvoCahDoe/invoice-layout-yolo11m) |
| 6 | `yolov8m` | 0.8388 | 0.5289 | 0.9077 | 0.8580 | [YOLOv8m](https://huggingface.co/AvoCahDoe/invoice-layout-yolov8m) |

## License

Apache 2.0