Download model refactor #75
Conversation
| @@ -7,6 +7,9 @@ | |||
| from torch.backends import cudnn as cudnn | |||
| from tqdm import tqdm | |||
|
|
|||
| from rt_gene.download_tools import download_external_landmark_models | |||
| download_external_landmark_models() | |||
There was a problem hiding this comment.
Note that this cannot be moved in the __init__ as the SFDDetector import below requires the models to be existing.
There was a problem hiding this comment.
As you quite rightly thought - this is probably the least "nice" bit about this.
We have already modified the SFD detector to add in rospkg - could we not modify it further to make this less "hacky"?
|
We could further improve this by only downloading the models once they are required, but I think this adds quite a bit of complexity and would be prone to mistakes. |
| @@ -7,6 +7,9 @@ | |||
| from torch.backends import cudnn as cudnn | |||
| from tqdm import tqdm | |||
|
|
|||
| from rt_gene.download_tools import download_external_landmark_models | |||
| download_external_landmark_models() | |||
There was a problem hiding this comment.
As you quite rightly thought - this is probably the least "nice" bit about this.
We have already modified the SFD detector to add in rospkg - could we not modify it further to make this less "hacky"?
|
Alright, what do we think after the latest commit @ahmed-alhindawi? :) |
Fix #71