YOLOv12 MEDIUM Model

This is a YOLOv12 (You Only Look Once) model fine-tuned on a custom dataset.

Model Details

• Model Architecture: YOLOv12Medium
• Framework: Ultralytics YOLOv12
• Image Resolution: 640x640
• Training Epochs: 100
• Batch Size: 8

Classes

The model is trained to detect the following classes: drone

Usage

Using Ultralytics (Recommended)

from ultralytics import YOLO
from huggingface_hub import hf_hub_download

# Download the model
model_path = hf_hub_download(
    repo_id="rujutashashikanjoshi/yolo12-drone-detection-0205_full-100m",
    filename="best.pt"
)

# Load the model
model = YOLO(model_path)

# Run inference
results = model("your_image.jpg")

# Process results
for result in results:
    boxes = result.boxes  # Boxes object for bbox outputs
    print(f"Found {len(boxes)} detections")

Using the Model in a Pipeline

from ultralytics import YOLO

# Load from HuggingFace
model = YOLO("hf://rujutashashikanjoshi/yolo12-drone-detection-0205_full-100m/best.pt")

# Run inference
results = model.predict("image.jpg", conf=0.5)

# Display results
results[0].show()

Training Details

• Optimizer: SGD with momentum
• Learning Rate: Adaptive with cosine annealing
• Data Format: YOLO format (normalized xywh)
• Framework Version: ultralytics>=8.0.0

Files in this Repository

• best.pt: Best model checkpoint (highest mAP)
• last.pt: Last model checkpoint
• config.json: Model configuration
• class_names.txt: List of detection classes
• results.csv: Training metrics over epochs
• *.png: Training plots and visualizations

Performance Metrics

Training metrics are available in results.csv and visualized in the training plots.

License

This model is released under the AGPL-3.0 License.

Citation

@software{YOLOv12_ultralytics,
  author = {Jocher, Glenn},
  title = {Ultralytics YOLOv12},
  year = {2023},
  url = {https://github.com/ultralytics/ultralytics}
}