LoFTR
LoFTR copied to clipboard
Colab Weights downloading doesnt work
There seems to be some error when downloading the weights:
FileNotFoundError Traceback (most recent call last)
[<ipython-input-29-011ecbd77aeb>](https://localhost:8080/#) in <module>
6 matcher.load_state_dict(torch.load("weights/indoor_ds.ckpt")['state_dict'])
7 elif image_type == 'outdoor':
----> 8 matcher.load_state_dict(torch.load("weights/outdoor_ds.ckpt")['state_dict'])
9 else:
10 raise ValueError("Wrong image_type is given.")
2 frames
[/usr/local/lib/python3.8/dist-packages/torch/serialization.py](https://localhost:8080/#) in __init__(self, name, mode)
249 class _open_file(_opener):
250 def __init__(self, name, mode):
--> 251 super(_open_file, self).__init__(open(name, mode))
252
253 def __exit__(self, *args):
FileNotFoundError: [Errno 2] No such file or directory: 'weights/outdoor_ds.ckpt'
please can you give me this download way of "outdoor.ckpt"
Its just executing the example in https://colab.research.google.com/drive/1BgNIOjFHauFoNB95LGesHBIjioX74USW?usp=sharing
I guess you can download the weights file from this drive link, probably the old links are broken. https://drive.google.com/drive/folders/1xu2Pq6mZT5hmFgiYMBT9Zt8h1yO-3SIp
Wow,thanks !!!! you like save my life,because i want try it on my project in order to report for my teacher
You can substitute the code snipped to download the weights with the following:
# Download pretrained weights
!mkdir weights
%cd weights/
!gdown 19s3QvcCWQ6g-N1PrYlDCg-2mOJZ3kkgS # indoor-ds-new
!gdown 1kW2bQejjMlmE7FGberHrubXpE_ttX2LB # indoor-ds
!gdown 1M-VD35-qdB5Iw-AtbDBCKC7hPolFW9UY # outdoor-ds
!gdown 1p2lxEyQm4ykJXsSKgiJ6pGf9WxyqWOnC # outdoor-ot
%cd ..
I guess you can download the weights file from this drive link, probably the old links are broken. https://drive.google.com/drive/folders/1xu2Pq6mZT5hmFgiYMBT9Zt8h1yO-3SIp
Thank you for the weight files. However, with "indoor_ds_new.ckpt" the model doesn't work well. It only shows few matches while "outdoor_ds.ckpt" shows a lot of matches. And I used indoor images to test.
Maybe the model hasn't been trained enough or I should change some parameters?