---
title: "IMED2: Project"
date: 2021-11-25 11:00
categories: [Image S9, IMED2]
tags: [Image, S9, IMED2]
math: true
---
Lien de la [note Hackmd](https://hackmd.io/@lemasymasa/B1xHaAn_F)
# Part 1
## Retinal fundus image segmentation
![](https://i.imgur.com/jYTbkNu.png)
## Iterative Closest Point - ICP
![](https://i.imgur.com/6IVRysx.png)
Quand on itere, on reste bloques sur des minimums local
En remplacant points par courbes:
![](https://i.imgur.com/PScgsbe.png)
On construit un ensemble de courbes problables et on associe ces courbes probables
![](https://i.imgur.com/d6SNtCv.png)
Ca marche pas car on recale un objet 3D sur un objet 2D
![](https://i.imgur.com/wpcnOWY.png)
# Part 2
## Iterative Closest Point Registration
[Vedo](https://github.com/marcomusy/vedo)
[ICP](https://github.com/ClayFlanningan/icp)
![](https://i.imgur.com/Ge2l95H.png)
![](https://i.imgur.com/Jrn80tD.png)
# Project part 3
## Mosaicking application
[CVRL](https://projects.ics.forth.gr/cvrl/fire/)
![](https://i.imgur.com/aCe6CeB.png)
:::danger
A la fin, **ca ne marchera pas**
:::
Mais c'est pas grave, on aura essaye :)
## Pathology evolution
[CVRL](https://projects.ics.forth.gr/cvrl/fire/)
![](https://i.imgur.com/0rISO0w.png)
![](https://i.imgur.com/CvdxnQk.png)
:::info
On fait un suivi temporel d'une pathologie
:::