Lukas Hoyer

I'm not sure why the automatic download does not work. Here is the Google Drive link for a manual download: https://drive.google.com/file/d/1woRzEPVuhaafrS_2_GlsJuVRyxWaGO4O/view?usp=sharing Are you able to download it from that link?

You can find the relevant procedure for extracting the downloaded zip to the right paths here: https://github.com/lhoyer/improving_segmentation_with_selfsupervised_depth/blob/e6b602922ed280d748f876ee390c6e6a790cee17/models/utils.py#L108

Hi, Technically, HRDA can probably also be applied to tumor segmentation. For that purpose, you need to change the data loaders. You can have a look at https://github.com/lhoyer/HRDA/blob/master/configs/_base_/datasets/uda_gtaHR_to_cityscapesHR_1024x1024.py and https://github.com/lhoyer/HRDA/blob/master/mmseg/datasets/gta.py...

We are planning to add a license to both repos. I'll let you know as soon as it is ready.

Licenses are now available for HRDA and DAFormer.

@fuweifu-vtoo DAFormer can be trained on a single GPU within 16 hours. As this is a very reasonable training time, we opted for simplicity and did not use Multi-GPU training....

We have used a single Nvidia V100 for the training. It should be possible to run the code with an RTX 3090 as well. The size of the dataset does...

The flags `n_gpus` and `gpu_model` are for internal purposes only and have no functionality in this repository. I forgot to remove them. However, you can specify the GPU by setting...

Thank you for your interest in our work. I did not encounter this problem during training. I would suggest checking that `python tools/convert_datasets/gta.py` was run successfully and that the produced...

Hello, 1. These classes are chosen to be thing-classes (countable objects such as car or traffic light) in contrast to stuff-classes (background regions such as road). Please, also have a...