This document contains the key notes/points/script for the Milestone I presentation in regards to solution path. Use the final copy of the proposal document for reference.
Slide #1
Ultimate Deliverables
- Drone-mounted FPGA platform with interfaced camera
- Machine-learning implementation using an existing model (i.e. tiny-yolo)
- Pedestrian detection, detecting one or more pedestrians with bounding boxes -> Not the most important aspect!
- We will not be designing our own neural network, nor are we going to attempt to synthesize a large model using HLS
- Use of accelerators, likely facilitated through use of a soft-processor
- Drone-mounted transmitter & base-station
- All supplementary docs ([OPS DOC ESPECIALLY IMPORTANT])
Slide #2
Four important deliverable dates
Oct 15th - Proposal (Don’t need to include this in the slides)
Nov 25th - First Prototype
Tentative Implementation Deliverables:
- Initial demonstration of ML implementation and camera capture functionality
- If ahead of schedule, these will be combined into one demo.
Feb 10th - Second Prototype
Tentative Implementation Deliverables:
- Initial drone, control, transmission, and power systems demonstration
- Includes initial base-station implementation
- Continued ML/camera work demo
Apr 3rd - Final Product Delivery
Slide #3
Immediate Steps
Our approach to developing this platform is:
- Start Small - Since this is intended to be a research development platform to be built on, it better for us to have a simple, fully functional implementation rather than a large (yet buggy/incomplete) implementation [FOCUS ON LEGACY RISKS]. We will start with integrating MARLANN, and then duplicate/expand functionality as necessary.
- Start Quickly - We will be presenting the client with potential purchasing combinations next week, with orders sent out shortly after. We expect a large amount of toolchain and integration issues right off the bat, so we need to act fast to target Milestone II.
- Start Legally - A big risk is the legal compliance aspect - in particular drone and radio licensing. We are in contact with the UBC UAS team to get additional details regarding legal compliance, but it is expected that this will cause a large amount of overhead (ex. not being able to fly at UBC).