In the Hailo AI Software Suite January 2025 Release, I noticed that Real-ESRGAN – Super Resolution model was mentioned. However, I couldn’t find any comparison in terms of FPS and Power. Also, on the GitHub repository, the Super Resolution model was last updated two years ago without any further updates. Are there any current updates or improvements? I’m looking to use it in C++ with an expected FPS of 30. Could you provide guidance on this?
Hi @minh.vo,
You can find all the relevant information about the new model in the following documentation:
HAILO8 Super Resolution Model Documentation.
Additionally, this repository provides examples of how to use some of models. Please note that it doesn’t include all the release data.
If you have any questions, feel free to ask!
Best regards,
Ronit
Additionally to what @ronits suggested, for comparison between models you can check out Hailo’s model explorer:
1 Like
I’m working on developing an algorithm similar to the one showcased here, specifically for thermal cameras, and targeting deployment on the Hailo AI processor.
While BSD100 is a common benchmark dataset, it may not be ideal for this task. Do you have any suggestions for better-suited datasets or methodologies tailored to thermal imaging and edge AI processing? I’m particularly interested in approaches for real-time edge inferencing and optimizing for low power consumption, without compromising detection accuracy.
Would love to hear from the community about any relevant experiences, datasets, or pre-trained models that have worked well in similar scenarios! 