aatifjiwani / rgb-footprint-extract

A Semantic Segmentation Network for Urban-Scale Building Footprint Extraction Using RGB Satellite Imagery

Hello, welcome to my Github page :D here are some neato states

As a start, I've made 1178 commits, created 31 repos, contributed to 5 repos, and have made 80 pull requests.

Enjoy these colorful badges of the language statistics (I guess my notebooks overloaded everything)

rgb-footprint-extract's Issues

I want to ask a question about test,please~

About AICrowd dataset,train and val datasets have their corresponding annotations ,however,test have not.

After you train a model , what you use to test?

inference on single image?

Hi,

Thanks for the great work! Would you please suggest how to run inference on a single image? I would like to test the model on my own urban satellite data. Thanks!

sh test_deeplab.sh ，How to visualize test results？

URLError about class Tensor in evaluate file

Hi, I met a error called URLError when I use the class of evaluate file and the information shows below:
File "inference.py", line 84, in main
tester = Tester(args)
File "/root/autodl-tmp/models/deeplab/evaluate.py", line 47, in init
self.model = DeepLab(
File "/root/autodl-tmp/models/deeplab/modeling/deeplab.py", line 22, in init
self.backbone = build_backbone(backbone, output_stride, BatchNorm)
File "/root/autodl-tmp/models/deeplab/modeling/backbone/init.py", line 11, in build_backbone
return drn.drn_c_42(BatchNorm)
File "/root/autodl-tmp/models/deeplab/modeling/backbone/drn.py", line 324, in drn_c_42
pretrained = model_zoo.load_url(model_urls['drn-c-42'])
File "/root/miniconda3/lib/python3.8/site-packages/torch/hub.py", line 727, in load_state_dict_from_url
download_url_to_file(url, cached_file, hash_prefix, progress=progress)
File "/root/miniconda3/lib/python3.8/site-packages/torch/hub.py", line 593, in download_url_to_file
u = urlopen(req)
File "/root/miniconda3/lib/python3.8/urllib/request.py", line 222, in urlopen
return opener.open(url, data, timeout)
File "/root/miniconda3/lib/python3.8/urllib/request.py", line 525, in open
response = self._open(req, data)
File "/root/miniconda3/lib/python3.8/urllib/request.py", line 542, in _open
result = self._call_chain(self.handle_open, protocol, protocol +
File "/root/miniconda3/lib/python3.8/urllib/request.py", line 502, in _call_chain
result = func(*args)
File "/root/miniconda3/lib/python3.8/urllib/request.py", line 1383, in http_open
return self.do_open(http.client.HTTPConnection, req)
File "/root/miniconda3/lib/python3.8/urllib/request.py", line 1357, in do_open
raise URLError(err)
urllib.error.URLError: <urlopen error [Errno 113] No route to host>
It turns out to be that the pretrain of DRN couldn't be download from http://dl.yf.io/drn/, and I can't open this url, do you have any solutions?

testing model on a cpu-only machine

Hi :) i run the infer_deeplab.sh setting --no-cuda but i get this error:

RuntimeError: Attempting to deserialize object on a CUDA device but torch.cuda.is_available() is False. If you are running on a CPU-only machine, please use torch.load with map_location=torch.device('cpu') to map your storages to the CPU.

Can you help me? Thank you

I added map_location=torch.device('cpu') in torch.load in loader.py

About training loss,I used my own dataset whose format is the same as what crowdAI and ,however after several epoches ，The loss value becomes negativeral.can you help me ?Thank you very much！

Best model weights

Could you release the weights for the best model, i.e. the one described as, "Our final model is a DeepLavV3+ module with a Dilated ResNet C42 backbone trained using the F-Beta Measure + Exponentially Weighted Cross Entropy Loss (Beta = 0.5). We employ the cross-task training strategy only for Urban3D and SpaceNet"?

--num-classes

model_arg is not defined

The evaluate script does not run due to model_arg not being defined -- https://github.com/aatifjiwani/rgb-footprint-extract/blob/main/models/deeplab/evaluate.py#L59.

pretrained model weights assume using GPU

FYI -- the pretrained model weights don't work if you aren't using a GPU as the model is only wrapped in DataParallel when you have a GPU here -- https://github.com/aatifjiwani/rgb-footprint-extract/blob/main/models/deeplab/evaluate.py#L58

evaluate model on own data

Hello ! How could I evaluate a pre-trained model on my own data?

Unable to Download Weights

Hello,
I saw the research work and am looking forward to try it's results on my custom dataset but when I clone into github it's not allowing me to download the weights file.
ERROR : This repository is over its data quota. Account responsible for LFS bandwidth should purchase more data packs to restore access.
@aatifjiwani Please upload the weights file and share a link of drive or any other source