@AlexBlackAI

PhD student at University of Surrey CVSSP | Adobe Content Authenticity Initiative

 London, England
alexblck.com
Joined October 2018
  • Tweets27
  • Following95
  • Followers14
6Media
Alex Black retweeted
CVMP@CVMP2 Dec 2023
Congratulations to Kar Balan, @AlexBlackAI, Andrew Gilbert, Simon Jenni, Andy Parsons and @JCollomosse for winning the best full paper award! #CVMP2023
1
6
9
1,179
VADER - Visit our poster you will! Video fingerprinting at #ICCV2023 trace provenance of video clips online and highlight any manipulations. VADER: Video Retrieval Alignment Differencing. The dataset: ANAKIN! arxiv.org/abs/2303.13193 @ContentAuth @cvssp_research @AdobeResearch
2
3
1
11
2,045
See you in Paris! #ICCV2023
1
485
Content Provenance: To Authenticity and Beyond! 3 years in 1 poster covering our group's research into media provenance and it's use cases in authenticity, attribution and creative economy #CVPR2023 AC workshop. Catch our keynote at @CVPR workshop on visual copy detection 6/19
4
36
3,331
Happy to share that our paper “ARIA: Adversarially Robust Image Attribution for Content Provenance” (arxiv.org/abs/2202.12860) is published at the #CVPR2022 Workshop on Media Forensics 🕵️ So why does adversarially robust content provenance matter? A paper thread below! 🧵1/n
1
8
1
27
Big congratulations on the award - Best Paper at #CVMP2021 🎉🎊🥂 @AlexBlackAI @tuvbui @JCollomosse
CVSSP Research@cvssp_research7 Dec 2021
Join us at the 2PM session 'Gimme the data!' @CVMP. Alexander Black is presenting 'VPN: Video Provenance Network for Robust Content Attribution' @AlexBlackAI @tuvbui @JCollomosse @AdobeResearch @DECaDE_Hub #cvmp2021 #AI #video #data 📜PDF 👉 dl.acm.org/doi/pdf/10.1145/3…
4
6
Great talk! 👏👍 @AlexBlackAI
CVSSP Research@cvssp_research7 Dec 2021
Join us at the 2PM session 'Gimme the data!' @CVMP. Alexander Black is presenting 'VPN: Video Provenance Network for Robust Content Attribution' @AlexBlackAI @tuvbui @JCollomosse @AdobeResearch @DECaDE_Hub #cvmp2021 #AI #video #data 📜PDF 👉 dl.acm.org/doi/pdf/10.1145/3…
1
2
Join us at the 2PM session 'Gimme the data!' @CVMP. Alexander Black is presenting 'VPN: Video Provenance Network for Robust Content Attribution' @AlexBlackAI @tuvbui @JCollomosse @AdobeResearch @DECaDE_Hub #cvmp2021 #AI #video #data 📜PDF 👉 dl.acm.org/doi/pdf/10.1145/3…
2
1
4
Join us at the @CVMP demo session today and explore 'Video Provenance Network for Rebust Content Attribution'. Full paper presentation tomorrow! @AlexBlackAI @tuvbui @JCollomosse @cvssp_research @AdobeResearch
6
1
6
Excited to be back at @CVMP! Great to be a part of the organising committee again this year. Looking forward to two days of amazing talks and networking! #cvmp2021 @MarcoVolino @ArminMustafa @JCollomosse @tuvbui @AlexBlackAI
4
8
Fake news and misinformation are major societal problems that we're excited to address through our image and video provenance AI research – at @cvssp_research and with Adobe and the Content Authenticity Initiative (CAI). Details in-thread for our 2021 papers at ICCV, CVPR, CVMP
5
7
9
Today we wrapped ‘Deep Discoveries’, our AHRC project developing visual search to create a virtual national collection across our museums, archives and cultural heritage institutions. Congrats to Dipu Manandhar leading CV/ML. #DeepDiscoveries @cvssp_research @UkNatArcRes @V_and_A
4
9
Alex Black retweeted
AK
@_akhaliq16 Jun 2021
Compositional Sketch Search pdf: arxiv.org/pdf/2106.08009.pdf abs: arxiv.org/abs/2106.08009
1
2
28
Maybe I should just do a recap of all the awesome @streamlit Drawable Canvas apps for its first anniversary 🙂 "Compositional Sketch Search" - an algorithm for searching image collections using free-hand sketches 😳 App: share.streamlit.io/alexblck/… by @AlexBlackAI
1
4
17
Replying to @LoLEsportsStats
@LoLEsportsStats @Quickshot [1 / 3] Thoughts about Win Expectancy (WE): There is nothing shocking about @MisfitsGG winning 70% of games with WE of 49.2%. It's a bit less likely (~11%) than winning 50% of games (~24.5%), sure. But the key here is the sample size of just 10 games.
1
[2 / 3] In fact, if we plot probabilities of winning for WE between 30% and 70%, we can see that they are not spread too drastically. Better teams are still more likely to win, but note that eg. the probability of being at 6 wins is almost the same for teams with 50% and 70% WE.
1
[3 / 3] The only way to separate these graphs and determine the better team is, of course, to play more games. Drawing conclusions from small sample sizes can be misleading. This is obviously an undeniable scientific proof that @G2esports are going to win the split.
So I am now a #pytorchudacityscholar! Thanks to @facebook and @udacity for this opportunity. A long and exciting journey lies ahead.
3
20
Alex Black retweeted
Dmytrii S@ml_review16 Oct 2018
A Closer Look at Spatiotemporal Convolutions for Action Recognition By Facebook @ylecun Empirical study of different spatiotemporal conv for action recognition Proposed R(2+1)D CNN for factorised spatiotemporal conv PyTorch github.com/facebookresearch/… ArXiv arxiv.org/abs/1711.11248
27
51
Here is my new blog post on flow-based deep generative models. Different from GAN or VAE, these model explicitly learn the probability density function of the real data using normalizing flows. lilianweng.github.io/lil-log…

Flow Based Deep Generative Models

lilianweng.github.io
5
131
4
470
Load more