search

cnexah / VA-DepthNet

VA-DepthNet: A Variational Approach to Single Image Depth Prediction
MIT License
113 stars 5 forks source link

Why does the measurement number change with each run evaluation? #6

Open dasda-asd opened 1 year ago

dasda-asd commented 1 year ago

Run the command for the first time: python ./vadepthnet/eval.py ./configs/arguments_eval_nyu.txt result: Computing errors for 654 eval samples , post_process: False silog, abs_rel, log10, rms, sq_rel, log_rms, d1, d2, d3 37.1363, 0.5363, 0.1786, 1.2995, 0.8748, 0.4910, 0.3442, 0.6113, 0.8001

Run the command for the second time: python ./vadepthnet/eval.py ./configs/arguments_eval_nyu.txt result: Computing errors for 654 eval samples , post_process: False silog, abs_rel, log10, rms, sq_rel, log_rms, d1, d2, d3 36.3882, 0.4038, 0.1615, 1.2035, 0.5721, 0.4457, 0.3762, 0.6654, 0.8434

Can you help me with this problem?

cnexah commented 1 year ago

Hi, thank you for your information. The result looks bad. I wonder if the pre-trained model is loaded correctly. Could you provide all the output information? For example, does the printed information include "Loading checkpoint" or "No checkpoint found"

dasda-asd commented 1 year ago

Hi, thank you for your information. The result looks bad. I wonder if the pre-trained model is loaded correctly. Could you provide all the output information? For example, does the printed information include "Loading checkpoint" or "No checkpoint found"

The output information

D:\xiaohe\anaconda\envs\zoedepth\lib\site-packages\torch\functional.py:504: UserWarning: torch.meshgrid: in an upcoming release, it will be required to pass the indexing argument. (Triggered internally at C:\cb\pytorch_1000000000000\work\aten\src\ATen\native\TensorShape.cpp:3191.) return _VF.meshgrid(tensors, **kwargs) # type: ignore[attr-defined] == Total number of parameters: 261660777 == Total number of learning parameters: 261660777 == Model Initialized == No checkpoint found at '/cluster/work/cvl/celiuce/code/var/vadepthnet/vadepthnet_nyu.pth' 2%|█▍ | 12/654 [00:07<06:26, 1.66it/s]

cnexah commented 1 year ago

Thank you for your information. Yes, the path of the pre trained model is incorrect. Please download the pre trained model for NYU dataset here: https://drive.google.com/file/d/1gQQ_9awBhElHTlsVxgbfuy2540ZjlZWq/view

and change the path in the configs/arguments_eval_nyu.txt