XOHW video1

NNEURaMA

Marco Bacis - Serena Farina - Stefania Deligia

The context: Diabetic Retinopathy

The problem: Microaneurysms Detection

(1) http://www.reviewofophthalmology.com

The solution: Convolutional Neural Networks

Four Convolution:D= 3x3 S=1N= 8, 16, 32, 32

4 x 4x32 2 x 2 x329x 9x32

51 x 51 x3

49 x 49 x8 24 x 24 x8

22 x 22x16 11 x 11 x16

Three MaxPooling:D= 3x3, 2x2, 3x3S= 2

Fully Connected

Softmax:p(MA) and p(!MA)

The tools: FPGA

(1) http://www.improveeyesighthq.com/diabetic-retinopathy.html(2) https://www.xes-inc.com/embedded-technologies/xilinx-fpga-technology/

Thanks for your attention

[email protected]@[email protected]

Please, follow us:

https://twitter.com/nneurama

https://www.facebook.com/pg/NNEURaMAatNECST/

https://www.facebook.com/pg/NNEURaMAatNECST/

XOHW17_NNEURaMA_public