Structured prediction requires substantial training data. new paper introduces the first few-shot scene graph model with predicates as functions within a graph convolution framework, resulting in the first semantically & spatially interpretable model.
https://arxiv.org/pdf/1906.04876.pdf

✴️ @AI_Python_EN

Is it a good idea to train RL policies from raw pixels? Could visual priors about the world help RL? We just released the code of our Mid-Level Vision paper addressing these questions. Spoiler: using raw pixels doesn’t generalize! Play with the results at http://perceptual.actor

✴️ @AI_Python_EN

Interesting NBER paper on the history of industry investment in basic research. "The Changing Structure of American Innovation: Some Cautionary Remarks for Economic Growth", Ashish Arora, Sharon Belenzon, Andrea Patacconi, Jungkyu Suh

https://www.nber.org/papers/w25893

✴️ @AI_Python_EN

the first wide-coverage Minimalist Grammar parser! Read all about it https://stanojevic.github.io/papers/2019_ACL_MG_Wide_Coverage.pdf
The work will be presented at #ACL2019

✴️ @AI_Python_EN

According to one urban legend, statistics depends on the normal distribution and, since most data aren't normally distributed, statistics is invalid.

This myth is easily busted. Many data, including natural phenomena, are in fact normally distributed. Secondly, the normal (Gaussian) distribution is but one of more than three dozen used in statistics. These two books concisely review the ones used most often by statisticians:

- Handbook of Statistical Distributions (Krishnamoorthy)
- Statistical Distributions (Forbes et al.)

As Bayesian statistics becomes mainstream, a good understanding of probability may be more important than ever. I've been cracking the books and have found these three quite helpful:

- Introduction to Probability (Bertsekas and Tsitsiklis)
- Introduction to Probability Models (Ross)
- Essentials of Probability Theory for Statisticians (Proschan and Shaw)

These three provide a somewhat philosophical take on probability:

- Probability, Statistics and Truth (von Mises)
- Probability Theory: The Logic of Science (Jaynes)
- Uncertainty: The Soul of Modeling, Probability & Statistics (Briggs)

Lastly, I'd recommend The Improbability Principle by David Hand to anyone.

✴️ @AI_Python_EN

There is no machine learning for dummies. It is a fact that we have to accept it !
machine learning is an advanced topic that needs knowledge of math, optimization algorithms and programming constraints.

Poeple love to hear stories about AI and how powerful machine learning is. However, they give up as soon as they see the first math equation.

If you want it, work for it ! do not dream for it !

#AI

✴️ @AI_Python_EN

Join Top Experts in Machine Learning, Deep Learning, NLP, AI Engineering for up to four days in San Francisco, and accelerate your career in 2019. October 29 - November 1. 60% OFF Ends Soon: https://hubs.ly/H0jf4Cg0
✴️ @AI_Python_EN

#TapNGhost: Novel Attack Techniques against #Smartphone Touchscreens - essentially by putting you phone down on a hacked table with #NFC you can hijack what the user taps on the screen. Never using my phone on a table again 😅 #hack #exploit #security youtu.be/kmYCXH4ax-g

Face Recognition System Using FaceNet

https://github.com/KarthikBalakrishnan11/Face_Recognition_FaceNet

✴️ @AI_Python_EN

Google Open Sources TensorNetwork , A Library For Faster ML And Physics Tasks
https://bit.ly/2F9vFus

✴️ @AI_Python_EN

Here's a demo of image classification with the webcam by using #tensorflowjs, the entire code is being run in the browser!

#machinelearning in the browser is the next big frontier of AI, as the world's 50% population is now online.

#Google's #tensorflowjs makes it easier to train and deploy machine learning/deep learning models in the browser itself. No major installations required, just a browser and internet!

https://lnkd.in/gxZvTJj
#ai #datascience #machinelearners #deeplearning

✴️ @AI_Python_EN

1000x Faster Data Augmentation

#ComputerVision #MachineLearning #ArtificialIntelligence

http://bit.ly/31hpcH0
✴️ @AI_Python_EN

Mona Jalal: Our paper is now on CVF Website. Check it out and please stop by our poster tomorrow #CVPR #CVPR2019 http://openaccess.thecvf.com/content_CVPRW_2019/html/WiCV/Jalal_SIDOD_A_Synthetic_Image_Dataset_for_3D_Object_Pose_Recognition_CVPRW_2019_paper.html

✴️ @AI_Python_EN

Neurons inside BERT when it gets trained on “Hey Jude” over and over.

✴️ @AI_Python_EN

Divide and Conquer the Embedding Space for Metric Learning" at #CVPR2019
Paper and Code: https://bit.ly/dcesml


✴️ @AI_Python_EN

Sound of Pixels: a network learning correspondences between image regions and sound components by watching unlabeled videos. http://sound-of-pixels.csail.mit.edu/
Cool work by Antonio Torralba's group! #CVPR2019

✴️ @AI_Python_EN

Camera localization techniques for AR require persistent storage of digital 3D maps. But deep neural networks can reconstruct detailed images of scenes from such maps. Our solution keeps 3D maps confidential while accurately computing camera pose https://aka.ms/AA5bu2n #CVPR2019
✴️ @AI_Python_EN

paper on Hybrid Task Cascade for Instance Segmentation, ranking 1st in COCO 2018 Challenge Object Detection task.
Project page: http://mmlab.ie.cuhk.edu.hk/projects/HybridTaskCascade/
Code: https://github.com/open-mmlab/mmdetection

✴️ @AI_Python_EN

Dr. Andrew Fitzgibbon is an expert in 3D #computervision and graphics. Discover work on body- and hand-tracking for tech like Kinect and HoloLens and hear how research on dolphins helped build models for the human hand:
https://aka.ms/AA5b1q9 #CVPR2019
✴️ @AI_Python_EN

hierarchical localization paper won the visual localization challenge at #CVPR2019
Paper: https://arxiv.org/abs/1812.03506

✴️ @AI_Python_EN