search

med-air / EndoNeRF

Neural Rendering for Stereo 3D Reconstruction of Deformable Tissues in Robotic Surgery
https://med-air.github.io/EndoNeRF/
173 stars 14 forks source link

Can you provide a sample evaluation metric file? #3

Open peiran88 opened 1 year ago

peiran88 commented 1 year ago

I tried to get PSNR, SSIM and LPIPS using the D-NeRF Quantitatively evaluation file: metrics.ipynb but the result seems not correct. My result for the cutting_tissues_twice dataset is {'mse': 0.02320987731218338, 'psnr': 16.343271255493164, 'ssim': 0.7986437082290649, 'lpips': 0.36871644854545593} which seems not correct compared with the paper. Did I make a mistake? I used the ground truth images(with tools) and the estimated RGB images to calculate these metrics, is that correct?

If possible, can you provide a sample evaluation metric file like metrics.ipynb of D-NeRF ?

yuehaowang commented 1 year ago

I didn't use the 'metrics.ipynb' script in D-NeRF to evaluate our method as I need to follow the evaluation procedure in E-DSSR for a fair comparison. I uploaded the evaluation script (eval_rgb.py) and provided a brief instruction in readme. Please check it out and let me know if it works.

chensjtu commented 1 year ago

I have read the eval_rgb.py, is this eval file used for the final metric reported in the manuscript? All the frames except the first one are used for train and all the images are used for eval. If so, I will follow this setting.

yuehaowang commented 1 year ago

To be precise, all frames are used for training and all images except the first one are used for evaluation.