florence-2
OnceUponATime-florence2-captionssoa-full-florence2
Smithsonian Open Access Dataset with Florence-2 Caption
日本語はこちら
This dataset is made of soa-full.
soa-full is an CC-0 image dataset from Smithsonian Open Access. However, the dataset does not contain the image caption.
Therefore, we caption the images by Florence 2.
Usage
from datasets import load_dataset
dataset = load_dataset("aipicasso/soa-full-florence2")
Intended Use
Research Vision & Language
Develop text-to-image model or image-to-text model.… See the full description on the dataset page: https://huggingface.co/datasets/aipicasso/soa-full-florence2.danbooru2023-florence2-caption
Danbooru2023 - Florence2 Caption dataset
This dataset contains captions of danbooru2023 images generated by microsoft/Florence-2-large
I use original one with task token
Format
parquet:
key: the danbooru id of the image
parsed: parsed florence 2 output of the image
Stat
MORE_DETAILED_CAPTION
Entries: 7,438,449
Output Tokens (Min/Max/Mean/Median):
Flan T5 Tokenizer: 19/736/120/114
DFN CLIP Tokenizer: 19/826/108.7/103
Qwen2 Tokenizer:… See the full description on the dataset page: https://huggingface.co/datasets/KBlueLeaf/danbooru2023-florence2-caption.megalith-10m-florence2
Megalith-10M with Florence-2 Caption
日本語はこちら
This reposity is the supplymentary of Megalith-10M.
Megalith-10M is an CC-0 like image dataset. However, the dataset does not contain the image caption.
Therefore, we caption the images by Florence 2.
Usage
from datasets import load_dataset
dataset = load_dataset("aipicasso/megalith-10m-florence2")
How to get images
git lfs install
git clone https://huggingface.co/datasets/drawthingsai/megalith-10m… See the full description on the dataset page: https://huggingface.co/datasets/aipicasso/megalith-10m-florence2.florence2_datasetimagenet-captions-florence2l
ImageNet-1k captions (Florence-2-large) + CLIP ViT-L/14 text embeddings
One caption per ImageNet-1k image (train 1,281,167 + val 50,000) and its CLIP text embedding, made
for caption-conditioned latent-diffusion experiments (pcls-exp wave 5). No images are included.
How they were made
Captioner: florence-community/Florence-2-large (777M, native transformers>=4.57),
task <DETAILED_CAPTION>, greedy decoding, max_new_tokens=80, fp16.
Input crop: the centre square of… See the full description on the dataset page: https://huggingface.co/datasets/hmdliu/imagenet-captions-florence2l.