truthdotphd/vessel-detection

Vessel Detection Model

This model performs vessel detection using YOLOv11 architecture. Two versions are available:

• YOLOv11-nano (public)
• YOLOv11-xlarge (private, enterprise)

Model Description

The model is trained to detect vessels in maritime imagery. It uses the YOLOv11 architecture with improvements in detection accuracy and speed.

Performance

YOLOv11-nano (Public Version)

• mAP50: 0.474
• mAP50-95: 0.192
• Precision: 0.587
• Recall: 0.476

YOLOv11-xlarge (Enterprise Version)

• mAP50: 0.579
• mAP50-95: 0.318
• Precision: 0.612
• Recall: 0.571

For access to the enterprise version, please contact admin@truth.phd

Usage

pip install huggingface_hub ultralytics matplotlib

from huggingface_hub import hf_hub_download
from ultralytics import YOLO
import matplotlib.pyplot as plt

# Download the model file and a sample image from Hugging Face
repo_id = "truthdotphd/vessel-detection"
model_path = hf_hub_download(repo_id=repo_id, filename="model.pt")
image_path = hf_hub_download(repo_id=repo_id, filename="vessels.jpg")

# Load the model
model = YOLO(model_path)

# Use the model for inference
results = model(image_path)
plt.figure(figsize=(10, 10))
plt.imshow(cv2.cvtColor(img, cv2.COLOR_BGR2RGB))  # Convert BGR to RGB for matplotlib
plt.axis('off')
plt.show()

Limitations

• Performance may vary depending on image quality and lighting conditions
• Optimized for daytime maritime imagery
• Detection accuracy may decrease in crowded scenes

Training Data

The model was trained on a proprietary dataset of maritime vessel images. The dataset includes various vessel types under different weather and lighting conditions.

Training Procedure

• Architecture: YOLOv11
• Training Framework: Ultralytics
• Hardware: NVIDIA GPUs
• Optimization: AdamW optimizer

For enterprise solutions and access to YOLOv11-xlarge, please contact admin@truth.phd