packshotter/blueprint-dataset
Blueprint OCR Dataset (YOLO Format) Images resized to 50% resolution for faster training and smaller download size. Quick Start from huggingface_hub import hf_hub_download import zipfile # Download zip_path = hf_hub_download( repo_id="packshotter/blueprint-dataset", filename="blueprint-dataset.zip", repo_type="dataset" ) # Extract with zipfile.ZipFile(zip_path, 'r') as z: z.extractall('.') # Train with YOLOv8 from ultralytics import YOLO model =… See the full description on the dataset page: https://huggingface.co/datasets/packshotter/blueprint-dataset.
2100
Blueprint OCR Dataset (YOLO Format)
Images resized to 50% resolution for faster training and smaller download size.
Quick Start
from huggingface_hub import hf_hub_download
import zipfile
# Download
zip_path = hf_hub_download(
repo_id="packshotter/blueprint-dataset",
filename="blueprint-dataset.zip",
repo_type="dataset"
)
# Extract
with zipfile.ZipFile(zip_path, 'r') as z:
z.extractall('.')
# Train with YOLOv8
from ultralytics import YOLO
model = YOLO('yolov8n.pt')
model.train(data='ocr_yolo_half/data.yaml', epochs=100, imgsz=640)Statistics
23 Classes
Structure (after extraction)
ocr_yolo_half/
├── data.yaml # YOLO config
├── images/
│ ├── train/ # 9,018 images (1754x2481)
│ └── val/ # 1,129 images
└── labels/
├── train/ # YOLO format .txt
└── val/