PIFu: Pixel-Aligned Implicit Function for High-Resolution Clothed Human Digitization

Method achieves state-of-the-art performance on a public benchmark and outperforms the prior work for clothed human digitization from a single image.

https://shunsukesaito.github.io/PIFu/

Code: https://github.com/shunsukesaito/PIFu

Paper: https://arxiv.org/abs/1905.05172

TuiGAN: Learning Versatile Image-to-Image Translation with Two Unpaired Images


TuiGAN can be use for various computer vision tasks ranging from image style transfer to object transformation and appearance transformation.

Github: https://github.com/linjx-ustc1106/TuiGAN-PyTorch

Paper: https://arxiv.org/abs/2004.04634

Huawei announced that its TensorFlow and PyTorch-style MindSpore Deep Learning middleware is now open source

https://towardsdatascience.com/huaweis-mindspore-a-new-competitor-for-tensorflow-and-pytorch-d319deff2aec

Github: https://github.com/mindspore-ai/mindspore

Docs: https://www.mindspore.cn/docs/en/0.1.0-alpha/architecture.html

Official tutorials: https://www.mindspore.cn/en

Nevergrad, an evolutionary optimization platform, adds new key features

Facebook AI’s open source Python3 library for derivative-free and evolutionary optimization.

https://ai.facebook.com/blog/nevergrad-an-evolutionary-optimization-platform-adds-new-key-features/

GitHub: https://github.com/facebookresearch/nevergrad

Documentation: https://facebookresearch.github.io/nevergrad/index.html

EfficientDet from Google: Towards Scalable and Efficient Object Detection

A new family of scalable and efficient object detectors. EfficientDet achieves state-of-the-art accuracy while being up to 9x smaller and using significantly less computation compared to prior state-of-the-art detectors

https://ai.googleblog.com/2020/04/efficientdet-towards-scalable-and.html

Github: https://github.com/google/automl/tree/master/efficientdet

Paper: https://arxiv.org/abs/1911.09070

Tutorial: https://github.com/google/automl/blob/master/efficientdet/tutorial.ipynb

Local-Global Video-Text Interactions for Temporal Grounding

This paper addresses the problem of text-to-video temporal grounding, which aims to identify the time interval in a video semantically relevant to a text query

Github: https://github.com/JonghwanMun/LGI4temporalgrounding

Paper: https://arxiv.org/abs/2004.07514

Announcing PyCaret 1.0.0

An open source low-code machine learning library in Python. PyCaret allows you to go from preparing data to deploying models within seconds from your choice of notebook environment.

https://towardsdatascience.com/announcing-pycaret-an-open-source-low-code-machine-learning-library-in-python-4a1f1aad8d46

Habr RU : https://habr.com/ru/company/otus/blog/497770/

Github: https://github.com/pycaret/pycaret

Guide: https://pycaret.org/guide/

Neural Networks from Scratch - Coding a Layer

A beginner’s guide to understanding the inner workings of Deep Learning

https://morioh.com/p/fb1b9f5a52bc

Video Part 1: https://www.youtube.com/watch?v=Wo5dMEP_BbI

Video Part 2: https://www.youtube.com/watch?v=lGLto9Xd7bU

Transform and Tell: Entity-Aware News Image Captioning

End-to-end model which generates captions for images embedded in news articles.

Github: https://github.com/alasdairtran/transform-and-tell

Demo: https://transform-and-tell.ml/

Paper: https://arxiv.org/abs/2004.08070

Today, on April 22 is Earth day. It’s a right time to look at the climate issues in terms of data storage.
* 90% of all data was created in the last two years
* IoT, Big Data and AI are huge data creators
* 70% of all data stored is copy data2
* 70-80% of data is typically unstructured * 2018 HDD shipments = 869Eb
* 2023 HDD shipments = 2.6Zb
* In a normal DC, 1 watt of HDD consumption = 1 watt of cooling

What can everyone do for the ecology of our planet?

* Migrate suitable workloads to the cloud
* Collect, process and store less data; archive more to reduce carbon storage
* Use backup/archive instead of big data
* Leverage copy management tools
* If you must keep data for longer, use tape or cloud tape

Software solutions for backup, managing and recovering data help to move your data to the cloud and so you can take care of the environment. Commvault - leading experts in software-defined storage. Over 11 Exabytes of customer data are under Commvault management.

ResNeSt: Split-Attention Networks

Split-Attention Network, A New ResNet Variant. It significantly boosts the performance of downstream models such as Mask R-CNN, Cascade R-CNN and DeepLabV3.

Github: https://github.com/zhanghang1989/ResNeSt#pretrained-models

Paper: https://arxiv.org/abs/2004.08955v1

The Illustrated GPT-2 (Visualizing Transformer Language Models)

Visual explaining the inner-workings of transformers, and how they’ve evolved since the original paper

https://jalammar.github.io/illustrated-gpt2/

Habr ru: https://habr.com/ru/post/490842/

OpenAI Implementation: https://github.com/openai/gpt-2

Training with quantization noise for extreme model compression

Quant-Noise is a new technique to enable extreme compression of models that still deliver high performance when deployed in practical applications.

https://ai.facebook.com/blog/training-with-quantization-noise-for-extreme-model-compression/

Paper: https://arxiv.org/abs/2004.07320

GitHub: https://github.com/pytorch/fairseq/tree/master/examples/quant_noise

Building a Real Time Emotion Detection with Python

https://morioh.com/p/801c509dda99

Code: https://github.com/Dhanush45/Realtime-emotion-detectionusing-python

How I taught my computer to play Spot it! using OpenCV and Deep Learning

https://towardsdatascience.com/how-i-learned-my-computer-to-play-spot-it-using-opencv-and-deep-learning-ad1f017a3ec3

Habr ru: https://habr.com/ru/company/otus/blog/498800/

Code: https://github.com/henniedeharder/spotit/tree/master/DeepLearningSpotIt

Making Monolingual Sentence Embeddings Multilingual using Knowledge Distillation

Easy and efficient method to extend existing sentence embedding models to new languages. This allows to create multilingual versions from previously monolingual models.

Code: https://github.com/UKPLab/sentence-transformers

Paper: https://arxiv.org/abs/2004.09813v1

NVIDIA Announce MONAI Open Source AI Framework for Healthcare Research

MONAI is a PyTorch-based, open-source framework for deep learning in healthcare imaging.

https://blogs.nvidia.com/blog/2020/04/21/monai-open-source-framework-ai-healthcare/?ncid=so-twit-79443#cid=ix11_so-twit_en-us

Code: https://github.com/Project-MONAI/MONAI

Docs: https://monai.readthedocs.io/en/latest/

Google’s Optimizing Multiple Loss Functions with Loss-Conditional Training

https://ai.googleblog.com/2020/04/optimizing-multiple-loss-functions-with.html

Demo: https://sites.google.com/view/stochss/adjusting-parameters

ADJUSTABLE REAL-TIME STYLE TRANSFER: https://sites.google.com/view/stochss/adjusting-parameters