Does anyone have a good reproduction result？

CapsuleEndoscope / EndoSLAM

EndoSLAM Dataset and an Unsupervised Monocular Visual Odometry and Depth Estimation Approach for Endoscopic Videos: Endo-SfMLearner

https://data.mendeley.com/datasets/cd2rtzm23r/1

MIT License

235 stars 46 forks source link

Does anyone have a good reproduction result？ #24

Open lantudou opened 11 months ago

lantudou commented 11 months ago

Tried to train a model using sequence dataset with provided virtual data, and I found very poor performance and different loss recording in log. Could anyone give me more detail about training endoSfM model?

1484115316 commented 11 months ago

Can you teach me how to organize the dataset to train？

howardchina commented 7 months ago

Actually, the ground truth depth is not pixel-wise related to the RGB image. This could be the reason for your poor results.