# Project Brief Re-alignment Meeting

## Agenda
1. **Real time vs Non-real time**
* ??hardware issues force us to redirect the whole project application??;
* are we able to change this given that we have "real-time" for our project name??
* *which application strictly requires real-time*:
* real-time application:
* music
* gesture recognition
* assisted living?
* auslan
* non-real time application
* auslan (as a proof-of-concept)
* tennis
* action recognition (as a proof-of-concept)
* or radical idea?
3. Describe problem with current project brief + description
4. Compromise + Adjustments in our project brief
5. Break down our project brief into assignment 01 charter?

## Minutes

### Real time Issue
- We have a current issue hardware wise to implement a real time pose estimation system.
- We think that real time aspect is quite crucial to most application.
- Currently, out of our control if hardware is not powerful enough (VM or our local PCs).
- [ACTION] We can approach by focussing without real time FIRST, and then later try to re-integrated real time into the system if we have time/resources.

### Application of Real Time Pose Estimation
- [Tsz Kiu] There might be some hidden scope in other applications that we might want to do.
- [Yick] There are two components to our system
- Gesture Recognition
- Application of our system (Music / Auslan)
- [Yick] Proposing to focus more on the gesture recognition aspect of the system. Rather than focussing on forcing our application part of our system to perform real time tasks.
- [Yick] We can try to focus on an application (like Auslan) that retains it's value even with or without real time.
- [Tsz Kiu] We might be pushed back in schedule because we need to spend time again in researching.
- [Matt] Maybe we can change to PoseNet as it runs on the browser.

### Pros Vs Cons
### Sticking to Music Path
#### Pros
- We have done some research on music mapping.
- We have a defined goal from previous discussions.
- Tangible and measureable goal
- We have a system, that we aim to build
- Music synthesis is quite straight forward (from PureData or other software)

#### Cons
- Real Time issue with current hardware.

### Refocus on Gesture Recognition
#### Pros
- An application that retains its value with or without real time

#### Cons
- We need to rethink to get a defined goal.

## Aspects to think about in defining our project
- Problem Statement
- Project End Goal/Objectives
- Potential Issues and Benefits
- Deliverables

## Tsz Kiu's Input
- Two goals that we can potentially achieve in music
1. Develop plugins to program such as PureData or SuperCollider
- Connecting Pose Estimation software and using it's output to a gesture recognition system
- Plugin would describe the current gesture shown and how it maps to PureData/Supercolider
2. Interactive Music System development using SuperCollider
- Three Boxes: [Pose Estimation] -> [Gesture Recognition] -> [Sound Synthesis]
- Reference paper from Tsz Kiu can be used.

## Todo
- [Matt] Work on using PoseNet on the VM (using browser).
- [Team] Collaborate on email to Jonathan on insight for project direction.
- [Matt] Sending email to Lucas on help for VM.
- [Yick] Fill in project charter for both Music vs Auslan.