inference - Githubissues

fangchangma / self-supervised-depth-completion

ICRA 2019 "Self-supervised Sparse-to-Dense: Self-supervised Depth Completion from LiDAR and Monocular Camera"

MIT License

623 stars 135 forks source link

Open hello7623 opened 4 years ago

hello7623 commented 4 years ago

What should I do if I only want to use your code for a few images?

Rajat-Mehta commented 4 years ago

@fangchangma can you please provide us the instructions that we can follow to make inferences using your pretrained models?

@hello7623 or did you already figure that out? If yes, could you please tell how did you do that?

kfina91 commented 3 years ago

My findings, while looking around in the code and trying around:

there is no plain inference method.
You cannot use the pretrained model on only images. As far as I understand, the pretrained model uses the input 'd' and 'rgb', which means both sparse depth and rgb image data. Therefore you won't get around training the model yourself if you have no depth data.