Roadmap: Computer Vision

QR Scan

Target Detect

Implement QR Scan Functionality

QR Flight Communication Library

QR Flight Communication Implementation

QR Ground Communication Library

QR Ground Communication Implementation

Ingest new QR info from XBee

Display QR Info on Ground Station

Dataset Preparation

Train model on top-view-mutli-person

Train model on stanford drone dataset

Collect benchmark dataset

Benchmark top-view-mutli-person

Bechmark stanford drone

Retrain with selected datasets combined

Integrate trained weights into local YOLOv5

Scope piping processed video out from CV program

Setup Ground Station to Ingest Video

Setup ground station to display video

Setup video save and export to USB

Validate geolocation (UPDATE LATER)

Requirements to Tasks

2. Detect and high light intruder

a. Train the YoloV5 model for Aerial view of the person.

i. Prepare dataset for training (draw out the bounding box labels on the image)

ii. Prepare benchmark dataset

iii. Train and test each dataset independently
iv. Combine datasets and retrain, evaluate performance

v. extract the model weights and deploy the detection on the local machine.

vi. integrate with target acquisition code

b. Display video output to ground station

c. Export video to USB stick

d. Track the intruder for 500m:

i. implement the tracking module

e. Draw the intruder path on the map of the competition area

i. Get the map from Google Satellite

ii. Use the Geolocation module to get the sequence of lat and long data of the path

iii. Draw the path of the Google Satellite Image