Difference between implementation with Web UI. about clip-interrogator HOT 4 CLOSED

So I tested it and the base model is worse for most but better for a few. I can make it a simpler option to use it for anyone who wants to switch between the two.

from clip-interrogator.

Good question! I'll run both models through my benchmark suite and see which comes out with closer alignment to ViTL and ViTH

from clip-interrogator.

if you use the source directly now there is new option on Config object blip_model_type which can be set to "base" or "large". i expanded my benchmark image set and ran through both and find large outperforms base so i will keep large as default.

from clip-interrogator.

if you use the source directly now there is new option on Config object blip_model_type which can be set to "base" or "large". i expanded my benchmark image set and ran through both and find large outperforms base so i will keep large as default.如果你现在直接使用源代码，在配置对象 blip_model_type 上有一个新的选项，可以设置为“base”或“large”。我扩展了我的基准图像集，并通过两者运行，发现大优于基地，所以我将保持大作为默认值。

when i run "ci = Interrogator(Config(blip_model_type=None))", it show：
got an unexpected keyword argument 'blip_model_type'
how should i handle this?

from clip-interrogator.