Congratulations to Kaiming and ResNet. ResNet is 10 years old but Kaiming remains as young and brilliant as ever! We are deeply grateful for you and for everything you have done for computer vision, Kaiming (though he refuses to be on social media)
Interview w/ The Smartest Kid in the World: Deng Mingyang (邓明扬)
2nd year PhD student in MIT under He Kaiming (何恺明)
ICPC Gold medal 🥇 (1st place)
Putnam Fellow (Top 5)
IOI Gold Medal🥇 (1st place)
IMO Gold Medal🥇
Translated interview here: https://t.co/zKF3O826Lz
@jcjohnss@drfeifei@karpathy@cs231n Those were the good ole days! The lectures, assignments, and notes were amazing. Thank you all for putting together and freely sharing such a wonderful course. It had a big impact on my life.
@giffmana@generatorman_ai Feel like I used to be pretty good at this but it's more challenging now that author lists are much longer.
Jifeng Dai, Liang-Chieh Chen, Han Hu are a few names that would grab my attention
Is tracking task necessary for learning to track? Our new work on self-supervised correspondence learning shows it is not! By simply comparing two video frames, without negative pairs, the correspondence can emerge for tracking objects and pixels.
https://t.co/fmgCAiqWq4
(1/n)
Congrats Jie @jayleicn on being awarded the @AdobeResearch Fellowship!! 🥳 🎉 #ProudAdvisor
PS. Jie works on diverse aspects of vision+language research (esp. videos) & will be on the job market next year 🙂
(cc tlberg; @uncnlp)
https://t.co/sdG5cXcY76
https://t.co/Gklry77COs
"Snap looks to enhance augmented reality features with acquisition of British research lab Ariel AI"
For those not familiar with Ariel #AI, the startup includes Iasonas Kokkinos, George Papandreou, and Stefanos Zafeiriou 🇬🇷
https://t.co/LMf8oRH0xM
https://t.co/MZ9wbpRL2P
@abursuc Very important papers for distilling what is actually important for a problem. My least favorite papers propose a "new task" (not actually new) with weak baselines.
How to train very deep ConvNets without residual blocks? Our ICML paper on Deep Isometric Learning successfully trains 100-layer ConvNets without any shortcut connections nor normalization layers (BN/GN) on ImageNet.
Paper: https://t.co/mxMJO1Wub9
Code: https://t.co/8Omd8SYYS9
At @EveryPointIO we recently developed a real-time 3D reconstruction method that runs on an iPhone. At first we're measuring stockpiles to provide instant answers to @stockpilereport customers. Excited to explore other applications that can benefit from real-time 3D insights!