search

matsuren / crownconv360depth

360° Depth Estimation from Multiple Fisheye Images with Origami Crown Representation of Icosahedron (IROS2020)
MIT License
53 stars 7 forks source link

About pre-trained models #2

Closed Hyeonjoong-Jang closed 3 years ago

Hyeonjoong-Jang commented 3 years ago

Hi,

I appreciate for this nice work.

I was about to try your code, and I noticed that you have not provided any pre-trained models.

Could you please push one of the checkpoints you have?

Or at least could you let me know the required GPU memory or some more information about training?

Thank you.

Hyeonjoong-Jang commented 3 years ago

OK, I started to run the train.py code using a command

python train.py dataset/omnithings

For those who didn't notice, we should put depth_train_640 folder inside omnithings folder.

dataset_root |-- omnithings | |-- cam1 | |-- cam2 | |-- ... | |-- depth_train_640

But when I run the code, it stopped at a point that it didn't finish one epoch.

I'm figuring out.

image

------Solution------ In omnithings dataset I downloaded, there were some image files that have four-digit-length names like "0001.png" instead of "00001.png". So I had to handle it

matsuren commented 3 years ago

Thank you for your interest in our work!

In omnithings dataset I downloaded, there were some image files that have four-digit-length names like "0001.png" instead of "00001.png". So I had to handle it

Sorry about that. I forgot that I renamed them. I added instructions in README.md.

I will upload one of the pretrained models soon.

matsuren commented 3 years ago

Hi, I uploaded a pre-trained model here.
If you have any other questions, feel free to ask me.

Thanks,

matsuren commented 3 years ago

Closing. If you have any other questions, please re-open the issue.