search

jmliu206 / LIC_TCM

MIT License
159 stars 24 forks source link

Question about eval the pretrained model #25

Closed achel-x closed 3 days ago

achel-x commented 3 days ago

I download the pretrained models N=64 corresponds to 0.05.pth.tar, and N=128 to mse_lambda_0.05.pth.tar.

Unfortuantly, I ran the eval cmd at kodak24 test dataset. The results are weird.

If anyone has succefully ran the eval.py, could you please guide it to me?

/LIC_TCM$ python eval.py --checkpoint ./0.05.pth.tar --data '/home/hxxiong/dataset/kodac/' --cuda torch.cuda.is_available() True Loading ./0.05.pth.tar PSNR: 25.06dB MS-SSIM: 11.57dB Bit-rate: 1.289bpp PSNR: 25.48dB MS-SSIM: 11.85dB Bit-rate: 1.333bpp PSNR: 24.90dB MS-SSIM: 12.26dB Bit-rate: 1.346bpp PSNR: 27.81dB MS-SSIM: 14.99dB Bit-rate: 2.276bpp PSNR: 26.68dB MS-SSIM: 12.65dB Bit-rate: 1.478bpp PSNR: 25.56dB MS-SSIM: 14.02dB Bit-rate: 1.994bpp PSNR: 22.41dB MS-SSIM: 10.29dB Bit-rate: 1.089bpp PSNR: 24.53dB MS-SSIM: 11.15dB Bit-rate: 1.275bpp PSNR: 26.22dB MS-SSIM: 12.48dB Bit-rate: 1.680bpp PSNR: 25.28dB MS-SSIM: 12.96dB Bit-rate: 1.115bpp PSNR: 24.70dB MS-SSIM: 11.03dB Bit-rate: 1.628bpp PSNR: 27.80dB MS-SSIM: 15.87dB Bit-rate: 1.938bpp PSNR: 26.64dB MS-SSIM: 13.40dB Bit-rate: 1.577bpp PSNR: 25.70dB MS-SSIM: 12.72dB Bit-rate: 1.345bpp PSNR: 24.50dB MS-SSIM: 12.00dB Bit-rate: 1.574bpp PSNR: 23.56dB MS-SSIM: 10.94dB Bit-rate: 1.087bpp PSNR: 24.70dB MS-SSIM: 11.69dB Bit-rate: 2.317bpp PSNR: 26.01dB MS-SSIM: 13.41dB Bit-rate: 1.571bpp PSNR: 25.34dB MS-SSIM: 14.71dB Bit-rate: 1.895bpp PSNR: 25.73dB MS-SSIM: 10.29dB Bit-rate: 1.493bpp PSNR: 24.55dB MS-SSIM: 11.35dB Bit-rate: 1.324bpp PSNR: 24.39dB MS-SSIM: 11.51dB Bit-rate: 1.144bpp PSNR: 24.73dB MS-SSIM: 11.66dB Bit-rate: 1.338bpp PSNR: 25.98dB MS-SSIM: 12.72dB Bit-rate: 1.405bpp average_PSNR: 25.34dB average_MS-SSIM: 12.3969 average_Bit-rate: 1.521 bpp average_time: 0.093 ms

================================================ $ python eval.py --checkpoint ./mse_lambda_0.05.pth.tar --data '/home/hxxiong/dataset/kodac/' --cuda --real torch.cuda.is_available() True Loading ./mse_lambda_0.05.pth.tar Bitrate: 5.475bpp MS-SSIM: 5.54dB PSNR: 17.40dB Bitrate: 5.877bpp MS-SSIM: 5.18dB PSNR: 17.21dB Bitrate: 5.883bpp MS-SSIM: 6.29dB PSNR: 17.14dB Bitrate: 6.687bpp MS-SSIM: 5.25dB PSNR: 14.33dB Bitrate: 5.615bpp MS-SSIM: 5.33dB PSNR: 16.89dB Bitrate: 6.637bpp MS-SSIM: 5.03dB PSNR: 14.12dB Bitrate: 4.515bpp MS-SSIM: 6.27dB PSNR: 19.71dB Bitrate: 5.784bpp MS-SSIM: 6.26dB PSNR: 19.24dB Bitrate: 6.017bpp MS-SSIM: 5.51dB PSNR: 15.95dB Bitrate: 4.866bpp MS-SSIM: 8.41dB PSNR: 20.83dB Bitrate: 5.556bpp MS-SSIM: 6.08dB PSNR: 18.83dB Bitrate: 5.302bpp MS-SSIM: 7.53dB PSNR: 16.87dB Bitrate: 6.139bpp MS-SSIM: 5.24dB PSNR: 15.12dB Bitrate: 5.419bpp MS-SSIM: 6.99dB PSNR: 19.31dB Bitrate: 6.487bpp MS-SSIM: 4.71dB PSNR: 15.23dB Bitrate: 4.825bpp MS-SSIM: 7.11dB PSNR: 20.32dB Bitrate: 6.706bpp MS-SSIM: 6.64dB PSNR: 17.89dB Bitrate: 5.156bpp MS-SSIM: 6.62dB PSNR: 17.36dB Bitrate: 4.803bpp MS-SSIM: 8.23dB PSNR: 17.63dB Bitrate: 6.393bpp MS-SSIM: 4.30dB PSNR: 17.47dB Bitrate: 5.158bpp MS-SSIM: 7.34dB PSNR: 22.71dB Bitrate: 4.432bpp MS-SSIM: 7.61dB PSNR: 21.40dB Bitrate: 5.699bpp MS-SSIM: 5.17dB PSNR: 15.98dB Bitrate: 5.428bpp MS-SSIM: 5.89dB PSNR: 17.43dB average_PSNR: 17.77dB average_MS-SSIM: 6.1893 average_Bit-rate: 5.619 bpp average_time: 0.219 ms

==========================================

I have no idea about this problem. Here is the version of my enviroment Python 3.9.18 timm 0.6.12 torch 1.12.1+cu113 compressai 1.2.6.dev0 /home/hxxiong/CompressAI

achel-x commented 3 days ago

the issue seems to solved, closed.