Video

[Webinar] How to Boost Accuracy & Speed in Satellite & Aerial Image Object Detection

Resource Featured Image

Delve into the critical challenges of small object detection in satellite and aerial imagery, a prevalent issue spanning sectors from agriculture and smart cities to environmental monitoring and defense. Often represented by just a few pixels, small objects introduce significant hurdles, such as scant detail for recognition and high susceptibility to background noise.

This webinar highlights innovative solutions to these challenges, spotlighting the cutting-edge YOLO-NAS-Sat model.

Key takeaways:

  • State-of-the-art models: Discover YOLO-NAS-Sat and see how it stands up against the well-known YOLOv8 in small object detection tasks.
  • Specialized architectures: Delve into YOLO-NAS-Sat’s design, from backbone adjustments tailored for small objects to its innovative U-Net-style decoder that preserves critical details.
  • Optimization strategies: Gain actionable insights from Deci’s experiences in fine-tuning models on the Dota 2.0 dataset, a vast collection of aerial images.
  • Rethinking evaluation metrics: Investigate the shortcomings of traditional evaluation metrics like COCO mAP for small object detection, and consider alternatives, including [email protected] and distance-based metrics.


Don’t miss this opportunity to elevate your understanding and capabilities in small object detection. Watch now.

If you want to learn more about optimizing your computer vision applications, book a demo here.

Access the Webinar

Share
Add Your Heading Text Here
				
					from transformers import AutoFeatureExtractor, AutoModelForImageClassification

extractor = AutoFeatureExtractor.from_pretrained("microsoft/resnet-50")

model = AutoModelForImageClassification.from_pretrained("microsoft/resnet-50")