profile
viewpoint
Jason Yosinski yosinski Uber AI Labs; Recursion San Francisco, CA http://yosinski.com/

pyro-ppl/pyro 6608

Deep universal probabilistic programming with Python and PyTorch

greydanus/hamiltonian-nn 209

Code for our paper "Hamiltonian Neural Networks"

yosinski/convnet_transfer 127

Code for paper "How transferable are features in deep neural networks?"

goodfeli/galatea 94

This was my private research codebase during grad school. After graduation I made it all public

YixuanLi/convergent_learning 77

Code for paper "Convergent Learning: Do different neural networks learn the same representations?"

yosinski/caffe 32

Caffe

yosinski/GitResultsManager 23

A simple snapshotting system for managing research results using Git.

yosinski/aracnaHardware 16

aracnaHardware

yosinski/Cornell-CS4670-2010 3

OpenCV based face detection and classificaiton for Nokia N900. Class project for Cornell's Computer Vision class (CS 4670).

startedyosinski/deep-visualization-toolbox

started time in 4 minutes

startedyosinski/deep-visualization-toolbox

started time in 26 minutes

fork kaimast/SILK-USENIXATC2019

Code for "SILK: Preventing Latency Spikes in Log-Structured Merge Key-Value Stores" published in USENIX ATC 19

fork in an hour

startedyosinski/convnet_transfer

started time in 2 hours

fork kaimast/rust-rocksdb

rust wrapper for rocksdb

fork in 2 hours

startedalshedivat/al-folio

started time in 2 hours

startedAnySoftKeyboard/AnySoftKeyboard

started time in 5 hours

startedosmandapp/OsmAnd

started time in 6 hours

startedyosinski/deep-visualization-toolbox

started time in 2 days

created repositoryrasmusbergpalm/nanograd

A miniscule implementation of reverse mode auto-differentiation

created time in 2 days

startedyosinski/deep-visualization-toolbox

started time in 3 days

startedyosinski/deep-visualization-toolbox

started time in 3 days

startednamin/nfv-benchmark

started time in 4 days

startedyosinski/deep-visualization-toolbox

started time in 4 days

startedyosinski/deep-visualization-toolbox

started time in 4 days

startedmyrrlyn/bitvec

started time in 5 days

startedgltf-rs/gltf

started time in 6 days

startedAtheMathmo/cpuprofiler

started time in 8 days

starteddermesser/leveldb-rs

started time in 8 days

startedyosinski/deep-visualization-toolbox

started time in 9 days

startedyosinski/deep-visualization-toolbox

started time in 11 days

startedyosinski/deep-visualization-toolbox

started time in 11 days

startedyosinski/deep-visualization-toolbox

started time in 11 days

startedyosinski/deep-visualization-toolbox

started time in 11 days

startedyosinski/deep-visualization-toolbox

started time in 11 days

startedyosinski/deep-visualization-toolbox

started time in 12 days

issue openeduber-research/jpeg2dct

Resizing DCT

Hi!

Thanks for such a cool work!

I use your library for my Pytorch based model. I use fixed-size images, transform them to DCT format using your library, and stream them using data iterator.

Now, I would like to change my data iterator. Instead of using fixed image inputs, I would like to dynamically change the image size at every iteration, not epoch.

Is applying bilinear interpolation directly on DCT output equivalent to applying bilinear interpolation to images, and afterwards applying DCT?

If there is no big difference, I would like to go for the 1st one, as doing it is very simple. If I have to go for 2nd, I need to modify my data iterator.

Thanks,

created time in 12 days

issue commentuber-research/jpeg2dct

Support image transformations

Speeding up training is relevant, and that's what I personally care about.

Dear @dreamflasher, encoding/decoding is not the part of the process speeding up the inference. DCT is already a compressed form. In order to compress the given image to an equivalent size that DCT has, you have to use a reasonable amount of convolutional layers which require heavy computational need. The main idea of this work is using already compressed image form to avoid 1st and 2nd blocks in ResNet which involves lots of Convolutional layers.

dreamflasher

comment created time in 12 days

startedyosinski/deep-visualization-toolbox

started time in 12 days

more