DRT 494 · 3(2-3-5) · Year/term 4/2

AI Edge Computing

การประมวลผลปัญญาประดิษฐ์ที่อุปกรณ์ปลายทาง

Progress

Notional hours: 150 h (online/self-study 75 · in class/lab 75)

Course description

Principles and constraints of edge AI; edge hardware; model conversion and ONNX; model compression and acceleration; measuring latency, throughput and power; planning on-board deployment.

Thai description

หลักการและข้อจำกัดของการประมวลผลปัญญาประดิษฐ์บนอุปกรณ์ปลายทาง ฮาร์ดแวร์สำหรับอุปกรณ์ปลายทาง การแปลงโมเดลและรูปแบบ ONNX การลดขนาดและเร่งความเร็วโมเดล การวัดความหน่วง อัตราการประมวลผล และการใช้พลังงาน การวางแผนนำระบบขึ้นใช้งานบนอากาศยาน

Description source: New draft for the curriculum committee

Course learning outcomes (CLO)

CLO	Outcome	PLO
CLO1	Explain constraints of edge AI	PLO4
CLO2	Convert and optimise models for edge devices	PLO4
CLO3	Measure and report performance on hardware	PLO4PLO5

Learning modules

1Edge AI for drones
Weeks 1–3 · 30 h2 KU

Online (before class) · 15 h

Study the assigned knowledge units in advance, review media and take the module quiz

In class / field · 15 h

Lab or field practice from worksheets with a safety checklist

Learning evidence: Checked worksheets and quiz results

2Model conversion and ONNX
Weeks 4–6 · 30 h2 KU

Online (before class) · 15 h

Study the assigned knowledge units in advance, review media and take the module quiz

Deploying AI models to edge devicesIn development
ONNX RuntimeTH

In class / field · 15 h

Lab or field practice from worksheets with a safety checklist

Learning evidence: Checked worksheets and quiz results

3Quantisation
Weeks 7–9 · 30 h2 KU

Online (before class) · 15 h

Study the assigned knowledge units in advance, review media and take the module quiz

Deploying AI models to edge devicesIn development
Edge ImpulseTH

In class / field · 15 h

Lab or field practice from worksheets with a safety checklist

Learning evidence: Checked worksheets and quiz results

4Edge hardware
Weeks 10–12 · 30 h2 KU

Online (before class) · 15 h

Study the assigned knowledge units in advance, review media and take the module quiz

Picamera2TH
jetson-inferenceTH

In class / field · 15 h

Lab or field practice from worksheets with a safety checklist

Learning evidence: Checked worksheets and quiz results

5Benchmark and deployment plan
Weeks 13–15 · 30 h1 KU

Online (before class) · 15 h

Study the assigned knowledge units in advance, review media and take the module quiz

Edge benchmarking and deployment planningTH

In class / field · 15 h

Lab or field practice from worksheets with a safety checklist

Learning evidence: Checked worksheets and quiz results

Assessment (draft)

Labs and worksheets	35%
Module quizzes	10%
Midterm examination	20%
Mini-project or practical exam	35%

Knowledge domain

D04 Sensors and embedded systems D11 Artificial intelligence and computer vision D12 Automation, robotics and swarms

Key references

Warden, P., & Situnayake, D. (2019). TinyML. O'Reilly.
ONNX Runtime. Documentation. link