CVPR18: Tutorial: Software Engineering in Computer Vision Systems

Organizers: David Doria
Description: Each year top computer vision researchers from around the world gather at CVPR to present and discuss recent developments and results from both academic and industrial efforts. Topics include things such as new procedures for train-ing neural networks, evaluation results on standard classifica-tion datasets, and demonstrations of applications enabled by the new algorithms. While the algorithms themselves are very exciting and critical components, when handed a new algo-rithm there is a non-trivial amount of work required to design, develop, run, and maintain a full software system around the algorithm. In fact, many teams in industry have software engi-neers, not computer vision experts, as many of their members. This workshop is intended to shine a light on this typically unpublicized part of the process with hopes to share best practices, expose common hurdles, and explain the complexity to computer vision researchers who may not be familiar with large system development.
Schedule: 0830 Accelerating Algorithm Development, Evaluation, and Deployment by Providing Frameworks, David Doria (HERE) 0900 Reproducibility – An Industrial Practice, Jan Ernst (Siemens) 0930 Scaling Active Learning for the Development of Imagery-Derived Maps, Ben Kadlec (Uber)
1030 Building Computer Vision Systems With Open Source Software, Matt Turek (Kitware)
1100 Transforming Research Code Into Robust Multiplatform Mobile Products, Stephen Miller (Fyusion)
1130 Transfer Learning: Data Curation, Training, and Deployment Strategies, Tim Franklin (Microsoft)
1200 Research to Prod: Large Scale Visual Recognition in the Cloud, Wei Xia (Amazon)

Видео CVPR18: Tutorial: Software Engineering in Computer Vision Systems канала ComputerVisionFoundation Videos