search

TRI-ML / vidar

Other
560 stars 67 forks source link

Saving self-calibration models. #17

Open YasuThompson opened 2 years ago

YasuThompson commented 2 years ago

Hi. I have several questions about saving models to a local folder.

1, When I run the self-calibration code, in which folders are trained models saved to? 2, Which part of config files should I adjust to changed the folders? 3, Could you point to the part of the code which saves trained models?

I am sorry for the effortless basic questions. But I am now confused how config files are processed.

YasuThompson commented 2 years ago

I actually found that I had only to add the following to at the bottom of for example ucm_euroc.yaml to enable saving models. 

checkpoint:
    folder: /data/vidar/checkpoints       # Local folder to store checkpoints
    save_code: True                       # Save repository folder as well
    keep_top: 5                           # How many checkpoints should be stored

And trained models are saved in directories like 2022-08-02_06h02m10s/models/001.ckpt But I want to know how I can visualize depth maps with the trained models.

YasuThompson commented 2 years ago

Inference code like this would be very much appreciated.

KOuldamer commented 2 years ago

@YasuThompson I am looking for Intrinsic calibration evaluation .

noskill commented 2 years ago

@KOuldamer i found relevant configs here https://github.com/TRI-ML/vidar/tree/main/configs/papers/selfcalib

isJHan commented 1 year ago

Inference code like this would be very much appreciated.

Hi, I'm a beginner in DL and there is a question. Whether normalization like (data-avg)/variance is required during inference? I find the input image is fed in network without normalization which is done in training step in the example you give here. So, I'm very confused! Thank you !