The different accuracy - Githubissues

BeileiCui / EndoDAC

[MICCAI'2024] EndoDAC: Efficient Adapting Foundation Model for Self-Supervised Depth Estimation from Any Endoscopic Camera

30 stars 2 forks source link

The different accuracy #8

Closed suxuanya closed 6 days ago

suxuanya commented 2 months ago

Hello, I according to the method of segmentation of https://github.com/ShuweiShao/AF-SfMLearner data set. In order to extract simple left view, I modified the/root/EndoDAC/datasets/scared_dataset py code. As shown in the figure below: But I loaded checkpoints on google driver and ran CUDA_VISIBLE_DEVICES=0 python evaluate_depth.py --data_path \

load_weights_folder '/ logs/endodac/models/weights_19' - eval_mono code after received is different from in the paper the accuracy of the excuse me why is this so?

BeileiCui commented 2 months ago

Hi I checked again, I may have uploaded a different checkpoint from the one I mentioned in the paper. Anyway, this one is a little better than the one in the paper.