x-CK-x / Dataset-Curation-Tool

A tool for downloading from public image boards (which allow scraping) / preview your images & tags / edit your images & tags. Additional tabs for downloading other desired code repositories as well as S.O.T.A. diffusion and auto-tag/caption models for your purposes. Custom datasets can be added!
GNU General Public License v3.0
32 stars 7 forks source link

Support for New Auto-Tagging/Captioning Model/s Available #16

Open x-CK-x opened 10 months ago

x-CK-x commented 10 months ago

Several new model/s are available; The following must be done:

x-CK-x commented 10 months ago
x-CK-x commented 10 months ago

the new models: https://huggingface.co/Thouph

x-CK-x commented 10 months ago

I need to add to the configs, information pertaining to the file type: image

x-CK-x commented 10 months ago

https://huggingface.co/Thouph/eva02-clip-vit-large-7704/tree/main with:

https://huggingface.co/Thouph/experimental-eva02-vit-large-448-adventurous/tree/main with:

https://huggingface.co/Thouph/eva02-vit-large-448-8046/tree/main with:

https://huggingface.co/Thouph/experimental_efficientnetv2_m_8038/tree/main with:

https://huggingface.co/Thouph/experimental_efficientnetv2_m_8035/tree/main with:

(Additional ViT Model/s potentially available soon as well)

x-CK-x commented 10 months ago

Working on a possible way to automatically install CUDA in the same manner that it is done with auto1111. The purpose of which would be to provide users the option of using their GPU.

x-CK-x commented 10 months ago

balanced was chosen for now amongst all the effNet options: https://huggingface.co/Thouph/experimental_efficientnetv2_m_8035/blob/main/model_balanced.pth

and the following two additional models will be excluded for the time being: https://huggingface.co/Thouph/experimental_efficientnetv2_m_8038/tree/main https://huggingface.co/Thouph/experimental-eva02-vit-large-448-adventurous/tree/main

x-CK-x commented 10 months ago

https://github.com/x-CK-x/Dataset-Curation-Tool/commit/6306c1b0a03cb8928b024261f0532335cc0bbcf6

x-CK-x commented 10 months ago

https://github.com/x-CK-x/Dataset-Curation-Tool/commit/5adc7aa22a3cc1d3861c522071d568bc6b280f54

x-CK-x commented 10 months ago

https://github.com/x-CK-x/Dataset-Curation-Tool/commit/8cb8d12a25335110a1f7a797311b46e002adbc04

x-CK-x commented 10 months ago

Current TODO:

Create a generic handler for all the auto-tag/caption model/s using the configs just created.

GPU support: detect if GPU is available when user tries to enable GPU mode, if no gpu is present; then set mode back to cpu and display to user.

x-CK-x commented 10 months ago

https://github.com/x-CK-x/Dataset-Curation-Tool/issues/32

x-CK-x commented 10 months ago

https://github.com/x-CK-x/Dataset-Curation-Tool/commit/33ab23ac56d1e47ea8bb68738cbb45159077f1d0

x-CK-x commented 10 months ago

https://github.com/x-CK-x/Dataset-Curation-Tool/commit/50255a6653201302274885ed466bf6ab6a41e588

x-CK-x commented 10 months ago

I need to make additional edits to the naming conventions of zack's model & tags file to reflect the consistencies of the new config file

x-CK-x commented 10 months ago

code refactored to the following:

                                   ------- Model_Manager
                                 /
                               /
                             /
GUI ------- Auto_Tag_Manager -------- Dataset_Manager
                    |        \
                    |          \
                    |            \
                    |              ------- Custom_File_Manager
                    |              
                    |              
                    |              
                     -------- Meta_Data_Extractor