This task focused on creating a dataset that illustrates a text to image model blind spots. By blind spots refers to the areas where the model fails when generating the images from the prompt given.
The model tested is Efficient-Large-Model/Sana_Sprint_0.6B_1024px_diffusers.
The model was loaded using the diffusers library in Python. Below is the code snippet used for loading:
import torch
from diffusers… See the full description on the dataset page:
https://huggingface.co/datasets/momondi/blind_spot_dataset.
