sipp11
/
handai-detector
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity
You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
17 Commits
2 Branches
0 Tags
131 KiB
 
Tree: c8987ecd7b
Branches Tags
${ item.name }
Create tag ${ searchTerm }
Create branch ${ searchTerm }
from 'c8987ecd7b'
${ noResults }
sipp11 c8987ecd7b
Updates 		5 years ago
.vscode Switch tracking to opencv - much better 5 years ago
examples Switch tracking to opencv - much better 5 years ago
src Updates 5 years ago
.gitignore Detect by YOLO and track by dlib 5 years ago
README.md YOLO img and vdo detector 5 years ago
pip.txt Initial commit 5 years ago

README.md

Handai aerial detector

We need to analyze road users in mid-block crossing area between Handai monorail station and Handai hospital.

Output we need

  • What, when and where are road users
  • focus on
    • pedestrain
      • with walking poles - STILL UNSURE if it's possible to detect
    • handicap (with wheelchair)
    • car
    • bus
    • cyclist

How?

  • I'm still debating if I should use YOLO or SSD for the run.

Challenge?

Object tracking algorithm from both opencv and dlib are pretty much the same. Good tracking capability, but they would lose track when there is a building or even a small pillar blocking the sight. Thus, we need to detect them all separately on each zone and find a way to link "object" that comes to the blocking zone and get through it the other way. I guess this should be able to get all the trace we need and connecting them all together. Yes, when there are tons of objects overlapping, things would get much tougher. But that is the issue for the other day.