Qualcomm AI Training Webinar

Day1

Opening Session

Opening Speech delivered by Qualcomm

9：30-9：35

5

Agenda Brief

9：35-9：40

5

1.0 Self-introduction by teachers and students

Teachers introduce themselves
Attendees give briefs of who they are and what they are working on.

9：40-10：00

20

1.1 Introduction to AI

1.1.1 AI: What? How? Where?

10:00 - 10:30 a.m.

30

1.1.2 Qualcomm AI

1.1.3 AI vs. Machine learning vs. Deep Learning

1.1.4 Different types of Machine Learning

1.1.5 Basic concepts in Machine Learning & Deep Learning

1.2 Quick Tour of Deep Learning

1.2.1 From ML to DL: What is deep learning?

10:30 - 11:10 a.m

20

1.2.2 Datasets of Deep learning: From public to custom

20

Tea break 11:10 - 11:20 a.m.

1.2.3 Infrastructure of Deep learning: From hardware to software

11:20 - 12:00 a.m

40

Lunch 12:00-1:00

Q&A, Open Discussion 13:00-13:30

1.2.4 History, Present and Future

1:30 - 2:30 p.m.

10

10

20

20

Tea break 2:30 - 2:40 p.m.

1.3.1 Review

Review

2:40-2:50p.m

10

1.3.2 Model Conversion and Demo

Hardware Preparation: AI Kit

2:50-3:00p.m

10

SoftWare Preparation: Part 1 SNPE SDK Development Environment Setup

3:00-3:25p.m

25

SoftWare Preparation: Part 2 SNPE Application Development Tools

3:25-3:45p.m

20

Tea break 3:45 - 3:55 p.m.

AI Demos - Object Detector Demo Converting Model

3:55-4:15p.m

20

AI Demos - Object Detector Run Demo

4:15-4:30p.m

15

1.3.3 Solve AI Problem Skills

Solve AI Problem Skills

4:30-4:40p.m

10

1.3.4 Homework, Q & A

Homework, Q & A

4:40~

-

Day2

2.1  Foundation of Deep Learning

2.1.1 Perceptron & Multilayer Perceptron

9:30 - 10:00 a.m.

10

10

2.1.2 Basic neuron layers

10:00 - 10:30 a.m.

20

2.1.3 Loss functions

10:30 - 11:00 a.m

15

2.1.4 Optimizer

20

10

2.1.5 Prevent Over-fitting in Deep Learning

11:00 - 12:00 a.m

20

30

Lunch 12:00-13:00

Q&A, Open Discussion 13:00-13:30

2.2 Building Deep Learning Model

2.2.1 Classic models

1:30 - 2:20 p.m

10

30

10

2.2.2 Hyper-parameters & Tuning Tricks

2:20 - 2:50 p.m

10

2.2.3 Fine-tune & Transfer Learnin

10

2.2.4 Data pre-process & Data augmentation

10

2.2.5 Hands-on - Keras_MNIST

2:50-3:00p.m

10

Tea break 3:00 - 3:15 p.m

2.3.1 Review

Review

3:15-3:55p.m

10

2.3.2 SNPE Training Part 2

An Image Classifiers Demo

10

SNPE Introduction

10

SNPE Workflow

10

Supported Chipsets / Supported Network Layers

3:55-4:40p.m

5

User-defined Operations (UDO) Workflow

30

Limitations

5

CPU vs GPU vs DSP

5

Tea break 4:40 - 4:50 p.m

Run SNPE on Linux Machine

4:50-5:20p.m

10

Building and Running the C++ Application on ARM Android

10

Thermal Measurement

5

Solve Problem with SNPE

5

2.3.3 Q & A

Q & A

5:20~

-

Day3

3.1 Getting Started with TensorFlow

3.1.1 TensorFlow Overview

9:30 - 10:30 a.m.

10

3.1.2 Low level API

10

3.1.3 Middle level API

10

3.1.4 High level API: Keras

10

20

3.2  Basic Knowledge of Object Detection

3.2.1 Overview

10:30 - 11:00 a.m

10

3.2.2 Performance metrics

20

3.2.3 Traditional methods for Object Detection

11:00 - 12:00 a.m

20

3.2.4 Two-stage detection

20

3.2.5 One-stage detection

20

Lunch 12:00-13:00

Q&A, Open Discussion 13:00-13:30

3.2.6 Hands-on Object Detection

1:30 - 2:30 p.m

60

3.3 SNPE Training Part 3

Review

2:30 - 3:00 p.m

5

SNPE Benchmarking

25

AI demos – Object Detector Demo
Part 1 : Get Camera Stream by Camera2
Part 2 : JNI Interface
Part 3 : Integrating SNPE

3:00 - 3:15 p.m

15

Tea break 3:15 - 3:25p.m

AI demos – Face Recognition

3:25 - 4:00 p.m

25

Solve SNPE Problem Skills

10

Q & A, Others

4:00 ~