Efficient Deep Learning for Computer Vision

CVPR Workshop 2022

CVPR 2022

New Orleans, Louisiana

June 20, 2022

09:00am - 17:00pm CDT

Best paper award:

Title: Hybrid Consistency Training with Prototype Adaptation for Few-Shot Learning

Authors: Meng Ye (SRI International), Xiao Lin (SRI International), Giedrius T Burachas (SRI International), Ajay Divakaran (SRI, USA), Yi Yao (SRI International)

Workshop Overview

Computer Vision has a long history of academic research, and recent advances in deep learning have provided significant improvements in the ability to understand visual content. As a result of these research advances on problems such as object classification, object detection, and image segmentation, there has been a rapid increase in the adoption of Computer Vision in industry; however, mainstream Computer Vision research has given little consideration to speed or computation time, and even less to constraints such as power/energy, memory footprint and model size. Nevertheless, addressing all of these metrics is essential if advances in Computer Vision are going to be widely available on mobile and AR/VR devices. The morning session of the workshop is to create a venue for a consideration of this new generation of problems that arise as Computer Vision meets mobile and AR/VR systems constraints. In the afternoon session, we will make sure we have a good balance between software, hardware, and model optimizations with an emphasis on training efficient neural networks with high performance computing architectures. Particular topics that will be covered:

  • Mobile and AR/VR Applications

    • Novel mobile and AR/VR applications using Computer Vision such as image processing (e.g. style transfer, body tracking, face tracking, depth estimation) and augmented reality

    • Learning efficient deep neural networks under memory and computation constraints for on-device applications

    • Multi-task learning and efficient multi-task models

    • On-device training, federated learning on device

  • Efficient Neural Network and Architecture Search

    • Compact and efficient neural network architecture (CNN, Transformer, or mixed) for mobile and AR/VR devices

    • Compact and efficient generative neural networks for mobile and AR/VR devices

    • Hardware (latency, energy) aware neural network architectures search, targeted for mobile and AR/VR devices

    • Efficient architecture search algorithm for different vision tasks (detection, segmentation, generative models, etc.)

    • Optimization for latency, accuracy and memory usage, as motivated by embedded devices.

  • Neural Network Compression, Quantization and Hardware Acceleration

    • Model compression (sparsification, binarization, quantization, pruning, thresholding and coding etc.) for efficient inference with deep networks and other ML models

    • Investigations into the processor architectures (CPU vs GPU vs DSP) that best support mobile applications.

    • Hardware accelerators to support Computer Vision on mobile and AR/VR platforms.

    • Low-precision training/inference & acceleration of deep neural networks on mobile devices

  • Data-Efficient Learning

    • Methods, algorithms, systems that improves data efficiency of deep learning

    • Semi-, weakly-, self-supervised learning for computer vision tasks

    • Zero-shot, few-shots learning, language-supervised learning

    • Open-set recognition, detection, and segmentation

  • Low Power Computer Vision Challenge

    • UVA Video Track: Track multiple moving objects in video captured by an unmanned aerial vehicle (UAV) at Purdue University with people throwing balls.

    • FPGA Detection Track: Perform object detection using a field programming gate array (FPGA) on Xilinx's Deep Learning Processing Unit.

Program Summary

This workshop will feature invited talks, selected paper publication, and panel discussion. See the program section for details.

Call for papers

Important dates

Workshop paper submission deadline: March 25, 2022

Notification to authors: April 1st, 2022

Camera ready deadline: April 8th, 2022 -- this is a firm deadline per CVPR workshop program.

Submission instructions

We are following the CVPR paper format: https://cvpr2022.thecvf.com/author-guidelines

We accept two forms of papers:

  • Long paper: Long papers should not exceed 8 pages. All the requirements are the same as the CVPR paper style guide. Long papers are for presenting mature works. A long paper should not only describe novel ideas but also have full experiments and analyses to support the proposed ideas.

  • Short paper: short papers should not exceed 4 pages. All other requirements are the same as the CVPR paper style guide. Short papers are for sharing early stage ideas. A short paper should describe novel ideas and have basic experiments to support the ideas. Comprehensive experiments and analyses, however, are not necessary.

Note the page limit includes figures and tables, in the CVPR style. Additional pages containing only cited references are allowed. Papers with more than 4 pages (excluding references) will be reviewed as long papers, and papers with more than 8 pages (excluding references) will be rejected without review.

Example submission paper with detailed instructions: ReviewTemplate.pdf

LaTeX/Word Templates: cvpr2022-author_kit-v1_1-1.zip

A paper should be submitted using the above templates. The length should match that intended for final publication.

Submission website

Please submit your papers through https://cmt3.research.microsoft.com/ECV2022.