With about 7000 attendees, the 6 days virtual conference on computer vision concluded a plethora of paper presentations, workshops and tutorials. From the breakthroughs on computer vision to open-sourcing datasets and projects, this conference was loaded with interesting topics and areas including autonomous driving, video sensing, action recognition, and much more.
We have already covered the topics and tutorials from day 1 and 2, i.e. June 14th and 15th. In this article, we have listed down all the important topics and tutorials that have been discussed from 16th June to 19th June.
This year, the conference witnessed a record of 1,470 research papers on computer vision accepted from 6,656 valid submissions. On June 16th, the IEEE CVPR 2020 announced its best paper awards in three categories. They are mentioned below-
1| Best Paper Award
Unsupervised Learning of Probably Symmetric Deformable 3D Objects from Images in the Wild
About: In this paper, the researchers, Shangzhe Wu, Christian Rupprecht and Andrea Vedaldi from the University of Oxford proposed a technique to learn 3D deformable object categorisation from raw single-view images with no external supervision. The technique is built on an autoencoder that factors each input image into depth, albedo, viewpoint, among others.
Read the paper here.
2| Best Student Paper Award
BSP-Net: Generating Compact Meshes via Binary Space Partitioning
About: Leading methods for learning generative models of shapes rely on implicit functions, and generate meshes only after expensive iso-surfacing routines. To overcome these challenges, the researchers from Google Research and Simon Fraser University were inspired by a classical spatial data structure from computer graphics known as Binary Space Partitioning (BSP) to facilitate 3D learning.
Read the paper here.
3| Best Student Paper Honorable Mention
DeepCap: Monocular Human Performance Capture Using Weak Supervision
About: Researchers from Stanford University, Max Planck Institute for Informatics and Saarland Informatics Campus proposed a deep learning approach for monocular dense human performance capture using weak supervision known as DeepCap. The model is a learning-based 3D human performance capture approach that jointly tracks the skeletal pose and the nonrigid surface deformations from monocular images.
Read the paper here.
Below we have listed down the tutorials that have been covered from 16th June to 19th June at the CVPR 2020.
1| Deep Learning and Multiple Drone Vision
About: The tutorial offered an overview of all the related topics of drone vision, such as drone localisation and world mapping, target detection, target tracking and 3D localisation. The tutorial consisted of 4 talks, which were an introduction to drone imaging, semantic world mapping and drone localisation, deep learning for target detection and 2D target tracking and 3D target localisation.
Know more here.
2| Vision Models for Emerging Media Tech and Their Impact on CV
About: The tutorial consisted of three hour-long talks that follow the book “Vision models for HDR and WCG imaging,” published by Elsevier in November 2019 in their computer vision and pattern recognition series. The topics included were the biological basis of vision, brightness and colour perception, vision models for HDR and WCG, and other such.
Know more here.
3| Image Retrieval in the Wild
About: This tutorial covered several important components of building an image retrieval system for real-world applications. The topics included here were approximate nearest neighbour search, how an algorithm is utilised in an online C2C marketplace app, a systematic review for heterogeneous person re-identification and other such.
Know more here.
4| Large Scale Holistic Video Understanding
About: Holistic Video Understanding is a joint project of the KU Leuven, University of Bonn, KIT, ETH, and the HVU team. In this tutorial, the researchers intended to put effort into introducing holistic video understanding as a new challenge in the computer vision field. This challenge focuses on the recognition of scenes, objects, actions, attributes, and events in the real world and user-generated videos.
Know more here.
5| Contactless Health Monitoring With AI
About: Contactless health monitoring is an emerging research topic in computer vision. This tutorial gives an overview of the latest developments and applications in the three fields, i.e. cameras, radio frequency and multi-modal sensing with an in-depth introduction on the core technologies invented by the group.
Know more here.
6| Automated ML Workflow for Distributed BigData Using Analytics Zoo
About: This tutorial presented how to implement the automated ML workflow for big data with a focus on supporting computer vision models and pipelines by seamlessly integrating different technologies including deep learning frameworks such as TensorFlow, Keras, PyTorch, etc., distributed analytics frameworks such as Apache Spark, Apache Flink, Apache Kafka, Ray, etc., and AutoML techniques such as hyperparameter optimisations.
Know more here.
7| Towards Annotation-Efficient Learning
About: Real-word computer vision applications often require models that are able to learn with few annotated examples, and continually adapt to new data without forgetting prior knowledge. The tutorial discussed one of the next big challenges in computer vision, that is to develop learning approaches that can address the essential shortcomings of the existing methods.
Watch the tutorial here: –
8| Fairness Accountability Transparency and Ethics and Computer Vision
About: In this tutorial, you will learn the Fairness, Accountability, Transparency, and Ethics (FATE) of modern computer vision. The tutorial also highlighted the research on uncovering and mitigating issues of unfair bias and historical discrimination that trained machine learning models to learn to mimic and propagate.
Know more here.
9| Visual Physics: The Interplay Between Physics and Computer Vision
About: This tutorial discussed an increasingly popular class of hybrid methods that blend physics and learning. They also discussed what visual physics is and how, over the past decade, the structured algorithms have been superseded by deep learning algorithms with superior performance.
Know more here.
Wrapping Up
Computer Vision and Pattern Recognition conference is one of the most popular events around the globe where computer vision experts and researchers gather to share their work and views on the trending techniques on various computer vision topics such as object detection, video understanding, visual recognition, among others.
CVPR 2020 also had Satya Nadella, the CEO of Microsoft, as one of its speakers where he talked about the journey of computer vision at Microsoft Research, 4D understanding, Hololens, Azure Konect, among others.
Watch the video below-
