VCQI

Installation

1. Clone the repository and install dependencies:

   git clone https://github.com/NotTwist/vcqi.git
   cd vcqi
   pip install -r requirements.txt

2. Visit the Releases tab and download the pretrained model weights vcqi.joblib.

Usage

Compute Metrics

Before running inference on VCQI, you need to compute metrics (PSNR, SSIM, Color and ID block 2) for your dataset. Use the following command to compute metrics:

For a Single Pair of Videos:

python compute_metrics.py \
    --gt_video /path/to/ground_truth_video.mp4 \
    --filter_video /path/to/colorized_video.mp4 \
    --metrics_list psnr ssim colorfulness multilayer_inc \
    --metric_path /path/to/output_metrics

For Multiple Videos:

python compute_metrics.py \
    --videos_path /path/to/videos \
    --methods method1 method2 \
    --metrics_list psnr ssim colorfulness multilayer_inc \
    --metric_path /path/to/output_metrics

Run Inference with a VCQI

After computing the metrics, you can run inference with VCQI using pretrained weights:

python inference.py \
    --metrics-dir /path/to/output_metrics \
    --model-path /path/to/vcqi.joblib