Page MenuHomePhabricator

GSP: EndoCV detection challenge
Open, NormalPublic

Description

Description
Reproduction of the polyp detection code used in the EndoCV2022 challenge:

Path to the code, readme and the paper: /home/m246i/drives/mount/E130-GSP/PaperData/2022YamlahiEtAl_Endocv22/detection

Path to the data: /home/m246i/drives/mount/E130-GSP/ProcessedData/2022YamlahiEtAl_EndoCV22

UPDATE:
The E130-GSP drive is slow so the data is moved to the E130-Projekte drive.
Path to the data: /E130-Projekte/endocv2022/2022YamlahiEtAl_Endocv22_paper

Requirements
GPU with 16GB memory.

Maximum participants per Community Day
2 or 3 participants

Event Timeline

amiiney triaged this task as Normal priority.May 5 2022, 3:36 PM
amiiney created this task.
seitela moved this task from Restricted Project Column to Restricted Project Column on the Restricted Project board.May 6 2022, 8:52 AM

I wanted to work on the reproduction, but the GSP-Networkdrives was super slow (kb/s) that I couldn't copy the files to my local machine :/

The first step of the GSP process is cleared: @j806e and @m700r managed to adapt the code, fix some path related bugs, install pytorch for different types of GPUs and train the the polyp detection models.

GSP step 2 will take place in the next community day: Run the inference with the training weights generated by @m700r and @j806e

The second step of the GSP process is clearer: @j806e and @m700r managed to reproduce the results of the models used in the EndoCV challenge.

Next step would be to update the readme with all the modifications in the code we did during the GSP process.