SONA: An Accelerator for Transform-Domain Neural Networks with Sparse-Orthogonal Weights.
Pierre AbillamaZichen FanYu ChenHyochan AnQirui ZhangSeungkyu ChoiDavid T. BlaauwDennis SylvesterHun-Seok KimPublished in: ASAP (2023)
Keyphrases
- transform domain
- neural network
- spatial domain
- orthogonal matching pursuit
- discrete cosine
- hadamard transform
- distributed video coding
- transform coefficients
- dct domain
- macroblock
- robust watermarking scheme
- discrete cosine transform
- artificial neural networks
- linear combination
- video codec
- bit plane
- reconstruction quality
- wyner ziv video coding
- compressive sampling
- blind watermarking
- frequency domain
- software suite
- image reconstruction
- motion vectors
- sparse representation
- spread spectrum
- image representation
- motion estimation