Four optimization techniques for Machine Learning Inference on Raspberry Pi SBC

Link to the article:
www.hackster.io/dmitrywat/fas...
Hardware in the video:
reTerminal
www.seeedstudio.com/ReTermina...
Blink Blink ICE Tower CPU Cooling Fan for Raspberry Pi (Support Pi 4)
www.seeedstudio.com/Blink-Bli...
Links to the materials mentioned in the video (tell me if I forgot any):
ARMNN GIthub
github.com/ARM-software/armnn
XNNPACK GIthub
github.com/google/XNNPACK
TF-MOT Toolkit Documentation
www.tensorflow.org/model_opti...
Smart Pruning: Improve Machine Learning Performance on Mobile
community.arm.com/developer/i...
Support FP32 computations using INT8 weights in TensorFlow Lite XNNPACK delegate
github.com/tensorflow/tensorf...
0:10 Why I made this video
1:12 Designing ...
2:17 Reducing ...
3:51 Optimized ...
8:17 ... inference
9:35 Don't set your Pi on Fire
Credits for the music:
Nostalgia Drive - A Nostalgic Synthwave / Chillwave / Retrowave mix
• Nostalgia Drive - A No...
ED-209 - Frequency
Photon - Cosmos
Credits for the artwork:
Awesome Pi On Fire Demo on Bare metal Raspberry Pi
• Pi On Fire Raspberry P...

[@exploring-electronic]

Thanks, it was helpful!

[@Hardwareai]

You're welcome!

[@ludmilamaslova3773]

Thank you for the information, everything is very well explained.

[@Hardwareai]

Glad it was helpful! :)

[@ola0x98]

Well explained, thank you

[@Hardwareai]

Glad you liked it

[@levbereggelezo]

Very good

[@Hardwareai]

Thanks!

[@MatheusFP104]

hi! This results from XNNPACK and ArmNN are using the CPU? Have you tried to delegate to GPU? I'm currently getting beating up trying to compiling ArmNN libraries and dependencies, just a nightmare!

[@Hardwareai]

Yes, it's CPU. For Raspberry Pi, at least at the time this video was made it was not possible to utilize the GPU.