Efficient inference of Image-Based Neural Network Models in Reconfigurable Systems With Pruning and Quantization
Jose Flich, Laura Medina, Izan Catalán, Carles Hernández, Andrea Bragagnolo, Fabrice Auzanneau, David Briand
-
SPS
IEEE Members: $11.00
Non-members: $15.00Length: 00:21:11
in recent work, we studied the phaseless PCA (low rank phase retrieval) problem and developed a provably correct and fast alternating minimization (AltMin) solution for it called AltMinLowRaP. in this work, we develop a modification of AltMinLowRaP, called AltMinLowRaP-Ptych, that is designed for reducing the sample complexity (number of measurements required for accurate recovery) for dynamic Fourier ptychographic imaging. Fourier ptychography is a computational imaging technique that enables high-resolution microscopy using multiple low-resolution cameras. Via exhaustive experiments on real image sequences with simulated ptychographic measurements, we show the power of our algorithm for reducing the number of samples required for accurate recovery.