A 100 random, high-res image set, pre-captioned for machine learning (LORA) training.
( 217 MB ZIP file, pre-captioned for 512x512, 768x768 and 1024x1024, with captions and base 2048x2048 images included)
Themes include:
People
Transportation
Landscapes
Food & Objects
Contents:
2048x2048 resolution upscaled base images without captions
1024x1024 resolution images with captions
768x768 resolution images with captions
512x512 resolution images with captions
All stand alone text captions
Additional Information:
Images have been generated in ComfyUI with SDXL Checkpoint "jib-mix-realistic-xl".
Link: https://civarchive.com/models/194768/jib-mix-realistic-xl
(Generation mostly with plain background for maximum object focus)
Captions have been added by WD14 captioning and include prompt 'l0r4' as a prompt keyword.
(replace this keyword to your liking)
This dataset has been used to train my "SD 15 Pixelized Art Lora".
https://civarchive.com/models/523529/sd-15-pixelized-art-lora-style
(Made by putting a pixelization and low res filter on the images of this dataset)
Please leave a like and comment if you used this for any (Lora) training.
I'd be very interested to view any content being generated from training this content :)