[BUG]: (Getting "TypeError: XXX got an unexpected keyword argument 'pretrained'" when attempting to train or infer on pretrained models in Solaris.)

[Joshw125]

Summary of the bug

I've been following along with the FOSS4g 2019 Solaris Tutorials, and have been able to run everything up until the selimsef model inferer (all three), which throws the following error: "TypeError: init() got an unexpected keyword argument 'pretrained'." I don't get the error when running the XDXD inferer. I also get the error when I try to train a custom model using the pipeline, and I'm struggling to figure out why it works on the one model but not on others. Has anyone else encountered this and figured it out? Thanks for any help!

Steps to reproduce the bug

Here is a link to my notebook (opened for viewing + commenting), with the outputs printed out up until the error I encountered.

https://colab.research.google.com/drive/1n8A8pK9BSgWV9SWguHCsp--25OD4ky0F?usp=sharing

Screenshots

Environment information

• Using Google Colab
• solaris version: 0.3.0
• python version: 3.6.7

[Servando1990]

Hi! I'm trying to reproduce this tutorial series as well. Concretely stuck with Fine-tuning. I want to reproduce it with another dataset (Inria https://project.inria.fr/aerialimagelabeling/)

but got the following problem: num_samples should be a positive integer value, but got num_samples=0. Which is a path error

I wonder how you have the csv train and infer files ?

Mine is :/content/drive/My Drive/solaris/solaris_tutorials/Solaris_FOSS4G_2019/data/austin/austin1.tif

[Joshw125]

Hello! I have my path set up similarly to yours it looks like, did you make sure to change the path in the config file to match where your images are for both training and inferring? And do you have the training images paired with their respective masks? I'm still wrapping my head around this a bit, but I believe the yml config needs to path to the csv where you'll list the training pairs in two columns labeled "image" and "label," where image is the training image, and label is the mask.

This picture shows my config file with the path to the .csv (pictured below), which then paths to the images and their training masks.

[Servando1990]

Thanks for the quick response!

I'm pretty sure that I have the paths correctly both in config and csv

Inria dataset already has its respect binarized masks. So skipped the part of generating masks

part where generaring masks

[zensenlon]

Summary of the bug

I've been following along with the FOSS4g 2019 Solaris Tutorials, and have been able to run everything up until the selimsef model inferer (all three), which throws the following error: "TypeError: init() got an unexpected keyword argument 'pretrained'." I don't get the error when running the XDXD inferer. I also get the error when I try to train a custom model using the pipeline, and I'm struggling to figure out why it works on the one model but not on others. Has anyone else encountered this and figured it out? Thanks for any help!

Steps to reproduce the bug

Here is a link to my notebook (opened for viewing + commenting), with the outputs printed out up until the error I encountered.

https://colab.research.google.com/drive/1n8A8pK9BSgWV9SWguHCsp--25OD4ky0F?usp=sharing

Screenshots

Environment information

• Using Google Colab
• solaris version: 0.3.0
• python version: 3.6.7

Hi, do you know how to deal with bug. I have the same problem.