Add document detection

Author: yushulx

examples/document-detection/CA01_01.jpg

@@ -0,0 +1,94 @@
+# YOLO Document Segmentation Training with MIDV500 Dataset
+
+This directory contains a complete pipeline for training YOLO segmentation models for document detection using the [MIDV500 dataset](https://github.com/fcakyon/midv500/tree/master) and [ultralytics](https://pypi.org/project/ultralytics/).
+
+## 🚀 Quick Start
+
+### 1. Install Dependencies
+```bash
+pip install -r requirements.txt
+```
+
+### 2. Prepare Dataset
+Download MIDV500 dataset and convert to YOLO format:
+```bash
+python data.py
+python prep_midv500_to_yolov11seg.py
+```
+
+### 3. Visualize Dataset
+Verify dataset quality with the GUI viewer:
+```bash
+python dataset_viewer.py
+```
+
+![YOLO dataset viewer](https://www.dynamsoft.com/codepool/img/2025/09/yolo-dataset-viewer.png)
+
+### 4. Train Model
+Train YOLO11 segmentation model using the provided script:
+```bash
+python train_yolo_doc_detection.py
+```
+
+![YOLO segmentation training](https://www.dynamsoft.com/codepool/img/2025/09/yolo-training.png)
+
+### 5. Run GUI Application
+Launch the document detection GUI:
+```bash
+python document_detector_gui.py
+```
+
+![ID detection with YOLO segmentation](https://www.dynamsoft.com/codepool/img/2025/09/document-id-yolo-segmentation.png)
+
+## GPU Setup for Faster Training
+
+### Prerequisites
+
+1. **NVIDIA GPU** with CUDA support (GTX 1060 or better recommended)
+2. **NVIDIA drivers** installed and up-to-date
+3. **CUDA toolkit** (automatically installed with PyTorch)
+
+### Check GPU Availability
+
+First, verify your GPU is detected:
+
+```bash
+# Check NVIDIA GPU and driver
+nvidia-smi
+
+# Check current PyTorch GPU support
+python -c "import torch; print('CUDA available:', torch.cuda.is_available())"
+```
+
+### Install GPU-Enabled PyTorch
+
+If CUDA is not available, install GPU-enabled PyTorch:
+
+```bash
+# Uninstall CPU-only PyTorch
+pip uninstall torch torchvision torchaudio -y
+
+# Install GPU-enabled PyTorch (CUDA 12.1)
+pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu121
+```
+
+For other CUDA versions, visit: https://pytorch.org/get-started/locally/
+
+### Verify GPU Setup
+
+```bash
+python -c "import torch; print('CUDA available:', torch.cuda.is_available()); print('GPU name:', torch.cuda.get_device_name(0) if torch.cuda.is_available() else 'N/A')"
+```
+
+### GPU Training Commands
+
+Once GPU is enabled, use `device=0` for GPU training:
+
+```bash
+# GPU training (recommended)
+yolo task=segment mode=train model=yolov8s-seg.pt data=dataset/doc.yaml imgsz=640 epochs=80 batch=16 device=0
+
+# CPU training (fallback)
+yolo task=segment mode=train model=yolov8s-seg.pt data=dataset/doc.yaml imgsz=640 epochs=80 batch=8 device=cpu
+```
+

examples/document-detection/README.md

@@ -0,0 +1,6 @@
+import midv500
+
+dataset_dir = 'midv500_data/'
+
+dataset_name = "all"
+midv500.download_dataset(dataset_dir, dataset_name)