evaluation

📊 Evaluation

We provide a standalone evaluation script that reports PSNR, SSIM, LPIPS, and Delta E 2000 between result images and their corresponding ground-truth images.
This script is useful for benchmarking external methods. To evaluate our method specifically, please refer to main.

To evaluate a set of result images against ground-truth images, use:

python evaluation.py \
    --result_dir /path/to/result/image/folder \
    --gt_dir /path/to/ground-truth/image/folder \
    --device gpu   # options: gpu or cpu

---

✉️ Inquiries

For inquiries related to the evaluation of other methods reported in the paper, please contact Ran Zhang (ran.zhang@samsung.com).