hustvl / ViTMatte

[Information Fusion (Vol.103, Mar. '24)] Boosting Image Matting with Pretrained Plain Vision Transformers
MIT License
339 stars 33 forks source link

VitTMatte pretrained models license #9

Closed leo27heady closed 1 year ago

leo27heady commented 1 year ago

Hello,

I came across the ViTMatte repository, and I'm interested in utilizing the image matting model for commercial purposes. However, I have a few questions regarding its usage and fine-tuning for my specific task. I noticed that there are two different implementations available - one trained on the Composition-1k dataset and the other on the Distinctions-646 dataset. I would appreciate some clarification regarding the following points:

  1. License: The repository is under the MIT license, but I would like to confirm whether this license permits the commercial usage of the model and if there are any additional restrictions or considerations to be aware of.

  2. Commercial Usage: Can I use the ViTMatte image matting model, as provided in the repository, for commercial purposes? Are there any specific conditions or requirements associated with commercial usage?

  3. Fine-tuning: I intend to fine-tune the pretrained ViTMatte model on my specific task or dataset. Does the model's license allow me to modify and fine-tune it for my commercial use case? Are there any specific guidelines or provisions for fine-tuning that I should follow?

Thank you for providing this valuable image matting model and for clarifying the terms of its usage. I look forward to your response.

JingfengYao commented 1 year ago

Thanks for liking ViTMatte. Our repo is totally under the MIT license. Feel free to use our model weights as long as you follow MIT license. I Hope ViTMatte could help. 😄

leo27heady commented 1 year ago

Thanks for the response!