Skip to content

Pseudocylindrical convolutions for Learned Omnidirectional Image Compression

License

Notifications You must be signed in to change notification settings

limuhit/pseudocylindrical_convolution

Repository files navigation

pseudocylindrical_convolution

Pseudocylindrical convolutions for Learned Omnidirectional Image Compression

Requirmed packages:

  • pytorch
  • cv2 (python-opencv)
  • numpy

Install:

  • python setup.py install
  • cd coder & python setup_linux.py install

Running the codec for 360-degree images:

  • Encoding:
    • python pseudo_codec.py --enc --img-file image_names.txt --code-file code_names.txt --model-idx 3 --ssim
    • python pseudo_codec.py --enc --img-list a.png b.png --code-list code_a code_b --model-idx 3 --ssim
  • Decoding:
    • python pseudo_codec.py --dec --out-file decoded_image_names.txt --code-file code_names.txt --model-idx 3 --ssim
    • python pseudo_codec.py --dec --out-list a_dec.png b_dec.png --code-list code_a code_b --model-idx 3 --ssim
  • Testing (Decoding and evaluate the performance):
    • python pseudo_codec.py --test --img-file source_image_names.txt --code-file code_names.txt --model-idx 3 --ssim
    • python pseudo_codec.py --test --img-list a.png b.png --code-list code_a code_b --model-idx 3 --ssim
    • python pseudo_codec.py --test --img-list a.png b.png --code-list code_a code_b --model-idx 3 --ssim

About

Pseudocylindrical convolutions for Learned Omnidirectional Image Compression

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published