Innovations in AI, Technology, and the Future Ahead

This week in AI & Machine Learning: Teaching dogs with ML, Azure in space, TensorFlow 3D, building object detection datasets, Swift for TensorFlow shuts down, and more

Artificial Intelligence News:

Nvidia: Researchers Train AI to Reward Dogs for Responding to Commands

A couple of researchers from Colorado State University used computer vision and a NVIDIA Jetson Nano to automatically reward a dog treats if it followed commands correctly. This fun project does a great job demonstrating the complexity of collecting data, data annotation, machine learning, and edge deployment.

Microsoft & Hewlett-Packard to Launch AI Capabilities to Space Station

Microsoft Azure’s cloud computing services and Hewlett-Packard Enterprise are attempting to deliver on artificial intelligence and edge computing capabilities to the International Space Station. The new machine learning capabilities can be used to help make predictions across a wide variety of problems that could help with space travel. We won’t be seeing HAL 9000 yet. 

$11.2 Billion Artificial Intelligence in Agriculture Markets

The artificial intelligence in agriculture shows no signs of slowing! With recent reports indicating it could be a 11.2 billion dollar industry by 2030. 

Developer Tools & Education:

3D Scene Understanding with TensorFlow 3D

We’re at an exciting time where machine learning libraries are making it easier to build 3 dimensional use case solutions. Read about how TensorFlow is bringing 3D capabilities to help understand data from Lidar, depth sensing cameras, and radar.  

Swift for TensorFlow Shuts Down

Development on Swift for TensorFlow officially stops. I know quite a few people were excited about the prospect of using Swift for machine learning, but it seems like that won’t be happening soon, at least not in the TensorFlow ecosystem. 

Uncovering Unknown Unknowns in Machine Learning

Discover some interesting weak spots of computer vision models, and join a global community of ML researchers and practitioners in a challenge to help better understand these weak spots.

Histogram matching with OpenCV, scikit-image, and Python

Learn how to perform histogram matching using OpenCV and scikit-image for style transfer. 

Hugging Face on PyTorch / XLA TPUs: Faster and cheaper training

This blog post provides an overview of changes made in the Hugging Face library, what the PyTorch / XLA library does, an example to get you started training your favorite transformers on Cloud TPUs and some performance benchmarks.

Interesting Podcasts & Interviews:

Building, Adopting, and Maturing LinkedIn’s Machine Learning Platform with Ya Xu | TWiML

Listen to Ya Zu of LinkedIn discuss experiences prior to becoming Head of Data Science, and how to build a sustainable machine learning platform. 

System Design for Autonomous Vehicles with Drago Anguelov | TWiML

Join this conversation with Drago Anguelov of Waymo self-driving cars to hear about autonomous vehicle system design, machine learning use cases, perception, planning, and much more. 

AI for Digital Health Innovation with Andrew Trister | TWiML

Hear Andrew Trister speak about the ways AI is transforming areas of healthcare innovation.

Notable Research Papers:

• Neural Re-rendering for Full-frame Video Stabilization

• A-NeRF: Surface-free Human 3D Pose Refinement via Neural Rendering

• Taming Transformers for High-Resolution Image Synthesis

• Unsupervised Semantic Segmentation by Contrasting Object Mask Proposals

• Few-Shot One-Class Classification via Meta-Learning

• Less is More: ClipBERT for Video-and-Language Learning via Sparse Sampling

• Multi-Task Reinforcement Learning with Context-based Representations

• Dynamic Neural Networks: A Survey

• Functional Space Analysis of Local GAN Convergence