Musings by Me(et)

• Semantic Segmentation using Fully Convolutional Networks over the years

  Jun 1, 2017 | semantic-segmentation, deep-learning, pytorch, visdom

  Fully Convolutional Networks (FCNs) are being used for semantic segmentation of natural images, for multi-modal medical image analysis and multispectral satellite image segmentation. Very similar to deep classification networks like VGG, ResNet, AlexNet etc there is also a large variety of deep architectures that perform semantic segmentation. I summarize several networks like FCN, SegNet, U-Net, RefineNet, PSPNet, G-FRNet etc here and provide reference Keras and PyTorch implementations for a number of them.

• 3D GAN - 3D Generative Adversarial Networks for Volume Generation and Classification

  Apr 1, 2017 | gan, deep-learning, tensorflow, visdom

  3D-GANs are 3-dimensional fully convolutional part of the GAN family. Published at NIPS'16, this architecture with it's adversarial critereon can generate near-perfect 3D Volumes and the discriminator features can be also used in 3D Volume classification to get state-of-the-art scores on the ShapeNet benchmark.

• Downloading datasets on AWS

  Dec 16, 2015 | aws, dataset, server

  If you happen to be working on huge datasets for Big Data or Machine Learning where the genome or image database is huge ( > 4 GB) and the data is available on a website only after user authentication, there are a lot of options to ponder over.

• Material + Jekyll

  Nov 25, 2015 | jekyll, material, webdev

  Was bored of editing every single `.html` file in my static site homepage and was very very difficult to manage the changes even with git. I had heard people saying cool stuff about Jekyll and its great static site generating power. And now that I have rolled this site up, I must say its magic :).

• Quartus-ModelSim on Linux

  Feb 17, 2015 | linux, vlsi

  Instruction Set to install and configure Quartus and Modelsim for Krypton Boards on 64 bit Linux systemss

subscribe via RSS