[Webinar] How to Speed Up YOLO Models on Snapdragon: Beyond Naive Quantization

Snapdragon is renowned for its high-quality graphics, efficiency, and optimized battery use, making it an excellent choice for various applications. However, optimizing object detection models, including those in the YOLO family, for real-time use on Snapdragon has been a challenge due to quantization issues. This session focuses on using the Snapdragon Neural Processing Engine (SNPE) to implement advanced INT8 quantization techniques, enabling faster model performance without losing accuracy.

Key Takeaways:

  • Advanced INT8 Quantization: Learn how to effectively apply INT8 quantization to improve model performance on Snapdragon.
  • YOLO Model Enhancements: See the performance improvements on YOLO models through before and after comparisons.
  • Automotive Applications: Learn how applying INT8 quantization to YOLO models running on Snapdragon can assist car manufacturers in enhancing their Advanced Driver-Assistance Systems (ADAS).
  • Expert Insights and Best Practices: Benefit from Lior and Eran’s knowledge and expertise as they share tips and best practices and answer your questions.

Watch now!

You May Also Like

[Webinar] How to Evaluate LLMs: Benchmarks, Vibe Checks, Judges, and Beyond

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

[Webinar] How to Efficiently Scale Video Analytics at the Edge

Add Your Heading Text Here
					from transformers import AutoFeatureExtractor, AutoModelForImageClassification

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

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