mindazhao / PWStableNet

Implementation of pixel-wise video stabilization
36 stars 13 forks source link

How to train with custom dataset? #3

Open zhaoxin111 opened 3 years ago

zhaoxin111 commented 3 years ago

Thank you very much for sharing such an excellent work. What data do I need to prepare to retrain the model? And there are main.py and main_new.py, which one should I use? Looking forward to your reply!

washingtonsk8 commented 3 years ago

Hello, @mindazhao! Thanks for sharing the code!

I have the same objective as @zhaoxin111 (btw, I believe main_new.py is the one, Zhao). My team and I intend to train it in a custom dataset. Can you provide us with a guide on creating the data for these folders in the parameters (image below)?

We would be grateful for that. Thank you.

image

RingoWRW commented 3 years ago

Hello, @mindazhao! Thanks for sharing the code! I have the same questions about training.

"parser.add_argument('--path_feature', default='../../feature_add/') ...

Thank you image

marcoosrs commented 2 years ago

Hello @mindazhao and everyone,

I would like to retrain the network as well, besides using a custom dataset. Did someone do this? Would it be possible to send me some additional code or instructions to generate the necessary information?

Thank you very much

Dalabg commented 1 year ago

Hello, @mindazhao! Thanks for sharing the code! I have the same questions about training.

"parser.add_argument('--path_feature', default='../../feature_add/') ...

Thank you image

Hello, my friend. Did you solve this problem? Please tell me how to do. Thanks!

Dalabg commented 1 year ago

Hello, @mindazhao! Thanks for sharing the code!

I have the same objective as @zhaoxin111 (btw, I believe main_new.py is the one, Zhao). My team and I intend to train it in a custom dataset. Can you provide us with a guide on creating the data for these folders in the parameters (image below)?

We would be grateful for that. Thank you.

image

Hello, my friend. Did you solve this problem? Please tell me how to do. Thanks!