Amy Zhao

Computer vision and machine learning researcher working on @Oculus at Facebook Reality Labs. PhD from @MIT_CSAIL.

[Adrian Dalca @AdrianDalca]

We're releasing a restructured #VoxelMorph registration (and much more) library at https://t.co/fNarjALqaI! Available in @pypi, support for both #tensorflow/#keras and #pytorch! Just in time for @MICCAI2020 :). @MIT_CSAIL @MIT @MGHMartinos @LCN @mit_caml
https://t.co/IgEdcTO1b1

[Adrian Dalca @AdrianDalca]

Tun in Thursday (July 23) @CHILconference -- @KatieLewisMIT will be presenting her work on registering clinical quality images (presentation video prerecorded - watch now! Q&A at 1:25 PM EST). with @nsanar and John Guttag @MIT_CSAIL @mit_caml ! Info: https://t.co/0WnOMexqrE

[Amy Zhao @AmyZhaoMIT]

Thanks @VentureBeat for the feature!

"MIT researchers train AI to predict how humans paint works of art: https://t.co/JdeQEZFvTx"

We are doing a live Q&A now for #CVPR2020 attendees! :) https://t.co/tTlXZE6hxS

[Amy Zhao @AmyZhaoMIT]

Excited for tomorrow's Q&A about my final research project from my PhD at MIT: synthesizing time lapses of paintings!

#CVPR2020

[Adrian Dalca @AdrianDalca]

@AmyZhaoMIT will be presenting her awesome work on Synthesizing Paintings of Time Lapse Videos on Wed 2-4pm PST. #cvpr2020 Q&A: https://t.co/ZWaMGAFIMK

Project: https://t.co/QQKnYJLtGm

with @balaguha, @KatieLewisMIT, @fredodurand, John Guttag, @mit_caml @MIT_CSAIL

[Amy Zhao @AmyZhaoMIT]

@dh7net @KatieLewisMIT @balaguha @AdrianDalca Thanks for your encouragement and advice on this work last year! It helped us along on this long journey :)

[Amy Zhao @AmyZhaoMIT]

Our work on synthesizing time lapses of paintings is now up on arXiv! What a fun and challenging project with @balaguha @KatieLewisMIT @AdrianDalca

https://t.co/AbH8YOPYcR

[DeepAI @DeepAI]

Painting Many Pasts: Synthesizing Time Lapse Videos of Paintings
https://t.co/nmKW3bmGsg
by Amy Zhao et al. including @AdrianDalca
#ConvolutionalNeuralNetworks #NeuralNetwork

[AK @_akhaliq]

Painting Many Pasts: Synthesizing Time Lapse Videos of Paintings
pdf: https://t.co/BX0IJSPm0e
abs: https://t.co/pR3Mf4SBJA
project page: https://t.co/PeifRtrxlc
video: https://t.co/E8hvrAf1Zu

[Amy Zhao @AmyZhaoMIT]

@crocodoyle @Hao_and_Y @drmnolan Love it!!! Yeah, really strong for that mana cost... Should be a rare :)

[Amy Zhao @AmyZhaoMIT]

Thank you @RSIPvision for the interview and the feature!

[MIT Clinical and Applied Machine Learning @mit_caml]

Amy continues to dominate the data augmentation game with a "Best of CVPR 2019" feature by @RSIPvision - congratulations to @AmyZhaoMIT! https://t.co/V3oShkqoO1

[Amy Zhao @AmyZhaoMIT]

Talking about learned data augmentation with co-authors Guha Balakrishnan and @AdrianDalca at poster 192 today!

#CVPR2019 #MRI

[Adrian Dalca @AdrianDalca]

Also check out @AmyZhaoMIT's feature in #CVPR2019 daily at https://t.co/892no9M9C5 @RSIPvision @mit_caml

[Amy Zhao @AmyZhaoMIT]

@GuilleJiCan We did some early experiments with GANs and found that they had a hard time synthesizing scans that are consistent with the labels (which is necessary for data augmentation). But it's possible that more experimentation could produce better results :)

[Amy Zhao @AmyZhaoMIT]

"One scan, no problem." A great article and interview by MIT news about our #CVPR2019 paper on data augmentation and one-shot #MRI segmentation! We talk about the method, its applications, and its fun roots in #MagicTheGathering.

[MIT Siegel Family Quest for Intelligence (SQI) @MIT_Quest]

“We’re hoping this will make image segmentation more accessible in realistic situations where you don’t have a lot of training data.” Don't miss @AmyZhaoMIT tomorrow at #CVPR2019. https://t.co/mYnKttNRTA

[Amy Zhao @AmyZhaoMIT]

@GuilleJiCan Thank you! Our intuition is that the learned transformations help to mimic the spatial and intensity variations in the population, while adding random transformations to the mix can help with overall robustness.

[Amy Zhao @AmyZhaoMIT]

Talking about Magic: The Gathering today, brain MRIs on Thursday! Data augmentation is useful everywhere :)

[MIT Clinical and Applied Machine Learning @mit_caml]

Check out @AmyZhaoMIT’s work at poster 160 during the #wicv workshop at #CVPR2019. She shows that by learning transformations to synthesize labeled data, you can improve few shot classification accuracy. She’ll also be talking about it during an oral on Thursday morning 💪🏾

[MIT Clinical and Applied Machine Learning @mit_caml]

@KatieLewisMIT is presenting her work at poster 147 during the #wicv workshop at #CVPR2019! Stop by to learn more about her method for faster, more accurate registration for MRI images 🧠

[Amy Zhao @AmyZhaoMIT]

@crocodoyle Not in this iteration unfortunately! Still dreaming of fitting it into a journal submission or something. I'd love to see the work published somewhere 🙂

[Amy Zhao @AmyZhaoMIT]

Our paper on learning how to do data augmentation (accepted to #cvpr2019) is up on arXiv! We show how to use our method to do medical image segmentation with just a single labeled example. We use publicly available datasets and our code is up on #github :) https://t.co/EwKKyaUB6Z

[Amy Zhao @AmyZhaoMIT]

@rbzhu Well it depends on your GPU memory but I would guess at least 2 datas

[MIT Clinical and Applied Machine Learning @mit_caml]

Encore presentation @ #ML4H by @KatieLewisMIT on SparseVM, a fast learning-based approach to clinical image presentation! Learn more at the poster session at 11:30, room 517D 🤓