Javier F.

Lightweight architectural designs of Convolutional Neural Networks (CNNs) together with quantization have paved the way for the deployment of demanding computer vision applications on mobile devices. Parallel to this, alternative formulations to the convolution operation such as FFT, Strassen and Winograd, have been adapted for use in CNNs offering further speedups. Winograd convolutions are the fastest known algorithm for spatially small convolutions, but exploiting their full potential comes… Show more

Lightweight architectural designs of Convolutional Neural Networks (CNNs) together with quantization have paved the way for the deployment of demanding computer vision applications on mobile devices. Parallel to this, alternative formulations to the convolution operation such as FFT, Strassen and Winograd, have been adapted for use in CNNs offering further speedups. Winograd convolutions are the fastest known algorithm for spatially small convolutions, but exploiting their full potential comes with the burden of numerical error, rendering them unusable in quantized contexts. In this work we propose a Winograd-aware formulation of convolution layers which exposes the numerical inaccuracies introduced by the Winograd transformations to the learning of the model parameters, enabling the design of competitive quantized models without impacting model size. We also address the source of the numerical error and propose a relaxation on the form of the transformation matrices, resulting in up to 10% higher classification accuracy on CIFAR-10. Finally, we propose wiNAS, a neural architecture search (NAS) framework that jointly optimizes a given macro-architecture for accuracy and latency leveraging Winograd-aware layers. A Winograd-aware ResNet-18 optimized with wiNAS for CIFAR-10 results in 2.66x speedup compared to im2row, one of the most widely used optimized convolution implementations, with no loss in accuracy. Show less

The geometry of 3D collagen networks is a key factor that influences the behavior of live cells within extracellular ma- trices. In this paper, we introduce a hybrid, two-step method for fully automatic quantification of the 3D collagen network geometry at fiber resolution in confocal reflection microscopy images. At first, a coarse binary mask of the entire network is obtained using steerable filtering and local Otsu thresholding. Second, individual collagen fibers are reconstructed by trac-… Show more

The geometry of 3D collagen networks is a key factor that influences the behavior of live cells within extracellular ma- trices. In this paper, we introduce a hybrid, two-step method for fully automatic quantification of the 3D collagen network geometry at fiber resolution in confocal reflection microscopy images. At first, a coarse binary mask of the entire network is obtained using steerable filtering and local Otsu thresholding. Second, individual collagen fibers are reconstructed by trac- ing maximum ridges in the Euclidean distance map of the bi- nary mask. The proposed method is validated on 3D collagen gels with various concentrations of collagen and Matrigel. We show that the presence of Matrigel in the gels affects the col- lagen network geometry by decreasing the network pore size while preserving the fiber length and fiber persistence length. Show less

The geometry of 3D collagen networks is a key factor that influences the behavior of live cells within extra- cellular matrices. This paper presents a method for automatic quantification of the 3D collagen network geometry with fiber resolution in confocal reflection microscopy images. The pro- posed method is based on a smoothing filter and binarization of the collagen network followed by a fiber reconstruction algorithm. The method is validated on 3D collagen gels with various collagen and… Show more

The geometry of 3D collagen networks is a key factor that influences the behavior of live cells within extra- cellular matrices. This paper presents a method for automatic quantification of the 3D collagen network geometry with fiber resolution in confocal reflection microscopy images. The pro- posed method is based on a smoothing filter and binarization of the collagen network followed by a fiber reconstruction algorithm. The method is validated on 3D collagen gels with various collagen and Matrigel concentrations. The results re- veal that Matrigel affects the collagen network geometry by decreasing the network pore size while preserving the fiber length and fiber persistence length. The influence of network composition and geometry, especially pore size, is preliminarily analyzed by quantifying the migration patterns of lung cancer cells within microfluidic devices filled with three different hydrogel types. The experiments reveal that Matrigel, while decreasing pore size, stimulates cell migration. Further studies on this relationship could be instrumental for the study of cancer metastasis and other biological processes involving cell migration. Show less