Weathered Vision Enhancing Object Detection in Adverse Weather
Authors
- Luv Surve
- Hitesh Bhardwaj
Dataset
This very-high-resolution (VHR) remote sensing image dataset was constructed by Dr. Gong Cheng et al. from Northwestern Polytechnical University (NWPU). It includes:
- 715 color images acquired from Google Earth with spatial resolution ranging from 0.5 to 2 meters.
- 85 pansharpened color infrared images acquired from Vaihingen data with a spatial resolution of 0.08 meters.
- In total, 800 VHR remote sensing images.
Each line of the corresponding text files defines a ground truth bounding box with the following coordinates:
(x1, y1) – [top-left coordinate]
(x2, y2) – [bottom-right coordinate]
The object classes are as follows:
- 1 – Airplane
- 2 – Ship
- 3 – Storage Tank
- 4 – Baseball Diamond
- 5 – Tennis Court
- 6 – Basketball Court
- 7 – Ground Track Field
- 8 – Harbor
- 9 – Bridge
- 10 – Vehicle
Model(s)
The model employed is YOLOv8 Medium (yolov8m.pt)
The YOLOv8 (You Only Look Once version 8) model is a state-of-the-art object detection model developed by Ultralytics. It is part of the YOLO (You Only Look Once) series, renowned for its speed and accuracy in real-time object detection tasks.
Version Information
- Model: Ultralytics YOLOv8.2.91 🚀
- Python Version: 3.10.12
Key Features
- Enhanced Performance: YOLOv8.2.91 represents a significant improvement over previous YOLO versions, offering superior accuracy and faster inference times.
- Real-Time Object Detection: The model is designed to detect objects in real-time, making it ideal for applications requiring immediate feedback, such as video surveillance and autonomous driving.
- High Efficiency: YOLOv8 is optimized for both speed and performance, utilizing advanced algorithms and optimizations to deliver high-quality results efficiently.
Results
We have the following results:
Results Examples
Table Test 1
Table Test 2
Results Examples
Here are some examples of the results:
Cloudy | Dense Cloudy | Foggy |
Night | Rainy | Snowy |