src.model.image_model module

Text-to-image model for generating synthetic images from prompts.

This module provides functionality to load and use diffusion-based text-to-image models for generating synthetic images based on text prompts and concepts.

class src.model.image_model.ImageModel(model_id: str, device: str, model_swapping: str, num_inference_steps: int, guidance_scale: int)[source]

Bases: Model

Text-to-image generation model wrapper.

Wraps diffusion-based text-to-image models to generate synthetic images from text prompts using specified configurations.

generate_images(n_images: int, prompt_text: str, concept: str) → Sequence[PIL.Image.Image][source]

Generate synthetic images for a given concept.

Generates multiple images by passing text prompts combined with a concept through the text-to-image model. Each image uses a different random seed for diversity.

Parameters:

  • n_images – Number of images to generate.

  • prompt_text – Base prompt text to use for generation.

  • concept – Concept name to include in the image prompt.

Returns:

Sequence of PIL Image objects generated from the prompts.