Generate an image from a given text prompt using a specified model,
then evaluate its quality and relevance.
This function:
1. Improves the input prompt using a Qwen3-based optimizer.
2. Generates an image with the chosen model.
3. Saves the image locally.
4. Evaluates the image on content consistency (SC_Score) and perceptual quality (PQ_Score).
5. Returns image path, filename, scores, evaluation remarks, and logs.
参数：
Args:
prompt (str):
The text description of the image to generate.
model_name (str):
The model used for image generation.
Must be one of:
- "black-forest-labs/FLUX.1-Krea-dev"
- "Qwen/Qwen-Image"
- "MusePublic/489_ckpt_FLUX_1"
- "MusePublic/flux-high-res"
Defaults to "black-forest-labs/FLUX.1-Krea-dev".
返回：
Returns:
Tuple[str, str, float, float, str, str, str]:
A tuple containing:
- save_path: Local path to the generated image (or placeholder if failed).
- fname: The filename of the image, or "Generation failed".
- sc: SC_Score (content consistency, 0.0–10.0).
- pq: PQ_Score (perceptual quality, 0.0–10.0).
- final_prompt: The optimized prompt used for generation.
- reason: Evaluation remarks or error messages.
- log_entry: Log entry with summary of results and evaluations.
Notes:
- If generation fails, a white placeholder image will be saved.
- Designed for Gradio MCP server (`gr.Image(type="filepath")` compatibility).
- Scores default to 0.0 if evaluation fails.
