Smart Adaptive Image Restoration CLI

This project restores low-quality images using Hugging Face image-to-image generation with automatic issue detection.

The workflow:

1. Load an input image
2. Analyze quality issues (blur, lighting, haze, lens-related artifacts)
3. Recommend prompt and generation settings
4. Run image generation with adaptive parameters
5. Save output image and process logs

Current project structure:

• cli.py: Main command-line entry point
• image_analyzer.py: Image quality analysis and adaptive recommendations
• image_to_image.py: Hugging Face generation logic
• process_logger.py: Per-run process logging setup
• requirements.txt: Python dependencies
• output/: Generated images are saved here by default
• my_logs/: Custom log folder used when running with --log-dir my_logs
• .venv/: Local virtual environment

Note:

• If you do not pass --log-dir, logs are saved in logs/ by default.

• Windows PowerShell (or any shell)
• Python 3.10+
• A Hugging Face token with access to the selected model/provider

1. Open terminal in project folder.
2. (Recommended) Create and activate virtual environment.
3. Install dependencies.
4. Set Hugging Face token in environment variable.

PowerShell commands:

python -m venv .venv
.\.venv\Scripts\Activate.ps1
pip install -r requirements.txt
$env:HF_TOKEN="your_hf_token_here"

If a virtual environment already exists, activate it and run only:

pip install -r requirements.txt
$env:HF_TOKEN="your_hf_token_here"

Interactive mode (recommended):

python cli.py

Interactive mode with custom log directory:

python cli.py --log-dir my_logs

Non-interactive mode (auto-suggested settings):

python cli.py --image "C:\path\to\image.jpg" --non-interactive

Optional arguments:

• --image: Input image path
• --prompt: Custom restoration prompt (overrides suggested prompt)
• --model: Model id (default is set in image_to_image.py)
• --output-dir: Output image directory (default: output)
• --log-dir: Log directory (default: logs)
• --non-interactive: Skip prompts and use analyzer suggestions

Output images:

• Saved to output/ by default
• Filename pattern: updatedimg_<input_name>_YYYYMMDD_HHMMSS.png

Log files:

• Saved to logs/ by default, or to custom folder from --log-dir
• Filename pattern: process_YYYYMMDD_HHMMSS.log
• Logs include run start, selected image, analysis summary, generation settings, output path, and errors

HF_TOKEN not set:

• Error indicates missing token
• Set it in current terminal:

$env:HF_TOKEN="your_hf_token_here"

Module not found errors:

• Activate your virtual environment and reinstall dependencies:

.\.venv\Scripts\Activate.ps1
pip install -r requirements.txt

Image file not found:

• Check full path and quotes around paths containing spaces

Model id validation errors:

• The code normalizes model ids to remove unsupported suffixes (for example provider routing suffixes)
• If access still fails, use a model available to your token/provider

• Denoising values above about 0.6 can alter identity/details significantly.
• For identity preservation (especially people/faces), keep denoising lower when possible.
• You can start with analyzer defaults, then fine-tune via interactive customization.