Mobile AI - Focus on BME Applications / Summer 2021

Updates

New Lecture is up: Lecture 4 | Detection [video] [slides]
Lecture on detection will be covered on Friday, September 3, 3pm at 801.
Dr Ahn asked to postpone his session to Monday, August 30, 3pm.
New Practicum released: [Practicum #3 - Detection]
New Practicum released: [Practicum #2 - Segmentation]
New Lecture is up: Guest Lecture by Dr Ahn | OpenCV in Android [video] [slides] [code]
📊 We announce leaderboard of teams! See more

Course Description

A 3-week intensive training course of end-to-end mobile AI: from training a deep learning model on custom data to its final deployment onto Android device.

What is Mobile AI?:
With the advance of computing powers on hardware devices, edge computing (see Wiki) started to break its own way. In contrast to cloud computing, edge computing is the paradigm which is closely associated with low latency, close to the point of service, or running locally. This shift towards local is especially useful for low-resource settings. With the rise of AI chips, built-in TPUs, prototyping AI applications becomes trendy, operating complex inferences to tackle computer vision and natural language processing tasks on-device.

Caveat:
In the following weeks, we will learn building Mobile AI applications with PyTorch + Android. This is arguably the first full course using PyTorch Mobile, which means a lot of things will be (re)engineered by me as an instructional material. Therefore, I look forward to your understanding, as well as feedbacks. Enjoy!
– San

Course Information

Time: Fridays, 3-5 pm
Venue: Building 110, Conference Room 801
Course repo (requires access): https://github.com/TBL-UNIST/mobile-ai-21

Prerequisites

Setups for deep learning:

conda environment:

conda create -n torchmobile python=3.8
conda activate torchmobile

PyTorch

pip install torch==1.9.0+cu111 torchvision==0.10.0+cu111 torchaudio==0.9.0 -f https://download.pytorch.org/whl/torch_stable.html

farabio
```
pip install farabio
```

Setups for mobile:

Download Android Studio first:
https://developer.android.com/studio
Download ‘삼성 USB 통합 드라이버’ - this makes AndroidStudio can detect your phone:
https://www.samsungsvc.co.kr/download
Unlock ‘Developer mode’ in your phone - this makes your phone permits access of AndroidStudio: https://meyouus.tistory.com/22
Please use these credentials for Google account:
email: unisttbl@gmail.com | password: tblXXXXXXXXXX

Previous Offerings

Summer 2020

Instructors

San Askaruly

Guest Lecturers

Myeong Hoon Choi

Dr. Yujin Ahn