Hundred-Kilobyte Lookup Tables for Efficient Single-Image Super-Resolution
This work addresses the challenge of efficient single-image super-resolution for edge AI devices constrained by power, computing, and storage resources, representing an incremental improvement over prior LUT-based methods.
The paper tackles the storage inefficiency of existing lookup table (LUT)-based super-resolution methods, which use multi-megabyte LUTs that hinder on-chip storage, by proposing hundred-kilobyte LUT (HKLUT) models that achieve uncompromising performance and superior hardware efficiency.
Conventional super-resolution (SR) schemes make heavy use of convolutional neural networks (CNNs), which involve intensive multiply-accumulate (MAC) operations, and require specialized hardware such as graphics processing units. This contradicts the regime of edge AI that often runs on devices strained by power, computing, and storage resources. Such a challenge has motivated a series of lookup table (LUT)-based SR schemes that employ simple LUT readout and largely elude CNN computation. Nonetheless, the multi-megabyte LUTs in existing methods still prohibit on-chip storage and necessitate off-chip memory transport. This work tackles this storage hurdle and innovates hundred-kilobyte LUT (HKLUT) models amenable to on-chip cache. Utilizing an asymmetric two-branch multistage network coupled with a suite of specialized kernel patterns, HKLUT demonstrates an uncompromising performance and superior hardware efficiency over existing LUT schemes. Our implementation is publicly available at: https://github.com/jasonli0707/hklut.