eldorado.tu-dortmund.de/server/api/core/bitstreams/4406a861-03fd-4a6a-89be-dbaac5149788/content
accuracy on mobile and embedded systems with strict resource limitations.
In our previous work in [9], we use a GPU to accelerate parts of our frequency domain analysis based classification, and compared [...] 29 µs per image for the Fourier features, and 17 µs for the Haar wavelet features. On an ARM Cortex A9 processor, the execution times are 276 and 88 µs, respectively. The classification performance is above [...] Core i7 4600U with 8 GB of random-access memory (RAM), the second platform, and on the ARM Cortex-A9 on the Zedboard, the third platform, to compare the execution time to the deepRacin implementation. …