Abstract
The SRNPU is the first ASIC implementation of the CNN-based SR algorithm which supports real-time Full-HD up-scaling and achieves higher restoration performance and power efficiency than previous SR hardware implementations. In this article, we propose an energy-efficient convolutional neural network (CNN) based super-resolution (SR) processor, super-resolution neural processing unit (SRNPU), for mobile applications. Traditionally, it is hard to realize real-time CNN-based SR on resource-limited platforms like mobile devices due to its massive amount of computation workload and communication bandwidth with external memory. The SRNPU can support the tile-based selective super-resolution (TSSR) which dynamically selects the proper sized CNN in a tile-by-tile manner. The TSSR reduces the computational workload of CNN-based SR by 31.1 % while maintaining image restoration performance. Moreover, a proposed selective caching based convolutional layer fusion (SC<sup>2</sup>LF) can reduce 78.8 % of external memory bandwidth with 93.2 % smaller on-chip memory footprint compared with previous layer fusion methods, by only caching short reuse distance intermediate feature maps. Additionally, reconfigurable cyclic ring architecture in the SRNPU enables maintaining high PE utilization by amortizing the reloading process caused by SC<sup>2</sup>LF operation under various convolutional layer configurations. The SRNPU is fabricated in 65 nm CMOS technology and occupies <inline-formula> <tex-math notation="LaTeX">$4 \times 4$ </tex-math></inline-formula> mm<sup>2</sup> die area. The SRNPU has a peak power efficiency of 1.9 TOPS/W at 0.75 V, 50 MHz. The SRNPU achieves 31.8 fps <inline-formula> <tex-math notation="LaTeX">$\times 2$ </tex-math></inline-formula> scale Full-HD generation and 88.3 fps <inline-formula> <tex-math notation="LaTeX">$\times 4$ </tex-math></inline-formula> scale Full-HD generation with higher restoration performance and power efficiency than previous SR hardware implementations. To the best of our knowledge, the SRNPU is the first ASIC implementation of the CNN-based SR algorithm which supports real-time Full-HD up-scaling.
Figure
figure 1
figure 2
figure 3
figure 4
figure 5
figure 6
figure 7
figure 8
figure 9
figure 10
figure 11
figure 12
figure 13
figure 14
figure 15
figure 16
Table
table I
table II
table III
table IV