Top 20 papers according to citation counter from academic.microsoft.com

https://www.kdnuggets.com/2018/03/top-20-deep-learning-papers-2018.html

#chart #dl

This is example how image can trick human brain, because blue lines are actually parallel.

So one shouldn’t blame neural networks for messing up classes.

High-level overview of different Object detection architectures: Fast R-CNN, Faster RCNN, Mask R-CNN, R-FCN, FPN, YOLO, SSD, RetinaNet:

https://medium.com/@jonathan_hui/what-do-we-learn-from-region-based-object-detectors-faster-r-cnn-r-fcn-fpn-7e354377a7c9

#dl #list #comparison #cv

Facebook is announcing a slew of new ML-related software projects and code releases this morning at F8, including Glow (a neural network compiler), PyTorch Translate (a public version of their production NMT code) and the roadmap for PyTorch 1.0 code.

https://code.facebook.com/posts/172423326753505/

#pytorch #facebook #deeplearning

CrowdHuman: A Benchmark for Detecting Human in a Crowd https://arxiv.org/abs/1805.00123

Andrew Ng announced that practical self-driving cars are ready. DriveAI will deploy a self-driving car service for public use in Texas starting in July.

Link: https://medium.com/@andrewng/self-driving-cars-are-here-aea1752b1ad0

#driving #rl #car #deeplearning #andrewng

Facebook’s Field Guide to Machine Learning video series

#guide #facebook

https://research.fb.com/the-facebook-field-guide-to-machine-learning-video-series/

Facebook report about advance in image recognition.

https://code.facebook.com/posts/1700437286678763/advancing-state-of-the-art-image-recognition-with-deep-learning-on-hashtags/

#cv #facebook #deeplearning

Open Neural Network Exchange — new format, introduced by Facebook to allow AI engineers to more easily move models between frameworks without having to do resource-intensive custom engineering. 

https://code.facebook.com/posts/1714175645317654/onnx-expansion-speeds-ai-development/

Neural network for extreme low-light photographs enhancement:

https://youtu.be/qWKUFK7MWvg

#CV #nn

Computer vision model training uses different augmentation of input images (random crops, distortions, etc.). Instead of hand-engineering the operation sequence, one can instead use meta learning to learn effective sequences. Gets new state-of-the-art of 83.54% top1 on ImageNet!

https://ai.googleblog.com/2018/06/improving-deep-learning-performance.html

Thomas Wolf’s post on how to make #Python #NLP faster.

Spoiler: use spaCy's internals and a bit of Cython magic

Bonus: a #Jupyter notebook with examples processing over 80 millions words per sec…

https://medium.com/huggingface/100-times-faster-natural-language-processing-in-python-ee32033bdced

ConvNet outperforms human dermatologists for melanoma detection.

Dermatologists in level-II protocol:
- sensitivity: 88.9% (±9.6%, P = 0.19)
- specificity: 75.7% (±11.7%, P < 0.05)

https://academic.oup.com/annonc/advance-article/doi/10.1093/annonc/mdy166/5004443

#cv #dl

Transfer Learning in Tensorflow (VGG19 on CIFAR-10): Part 1

Tutorial on how to use pretrained networks not to make them learn everything from scratch.

Link: https://towardsdatascience.com/transfer-learning-in-tensorflow-9e4f7eae3bb4

#tutorial #cv #tensorflow #dl

Introduction to numpy matrices syntax.

https://www.kdnuggets.com/2017/03/working-numpy-matrices.html

#numpy #Python #tutorial

Second part of the tutorial.

https://towardsdatascience.com/transfer-learning-in-tensorflow-5d2b6ad495cb

Brief intro into object detection with python.

https://towardsdatascience.com/object-detection-with-10-lines-of-code-d6cb4d86f606

#python #tensorflow #cv