IEEE Symposium on Evolving Deep and Transfer Learning Models for Computer Vision and Medical Imaging (IEEE ECV)

IEEE SSCI 2020 will run as a virtual conference.

Scope and Motivation

Automated diagnostic imaging problems are challenging owing to data scarcity, poor data quality (e.g. low contrast, occlusions, and distractors), complex characteristics of the diagnostic problems and subtle and delicate distinctiveness between benign and tumour scenarios. Deep learning and transfer learning show superior capabilities of tackling computer vision and automated medical diagnostic problems. Examples include the proposal and adoption of a variety of deep architectures for image synthesis (e.g. auto-encoders and Generative Adversarial Networks), segmentation (e.g. SegNet and Mask R-CNN), detection (e.g. YOLOv3), and classification (e.g. VGGNet, ResNet, ResNeXt, and SqueezeNet). Moreover, the transfer learning process based on pre-trained models is able to overcome barriers related to data scarcity by transferring learned features to a new task. It enables the networks to not only embed rich features learned from a wide range of non-medical images during pre-training, but also acquire new feature representations from the learning process of a new domain.

However, the design of new and effective deep learning models and identification of the optimal hyper-parameters of the resulting as well as transfer learning models require profound domain knowledge, which may not always be available to researchers. In parallel, evolutionary algorithms show powerful search capabilities of solving single-, multi-, and many-objective optimization problems. In this regard, the superior search capabilities of evolutionary computing algorithms allow them to tackle such optimization problems, e.g. to devise evolving deep neural networks that fit the tasks at hand, as well as to identify optimal hyper-parameters of the transfer learning process.

This special session aims to stimulate studies pertaining to not only complex deep learning-based computer vision and medical imaging systems but also optimal topology and hyper-parameter identification for such deep networks through evolutionary computing and related paradigms

Topics

Image segmentation & visual saliency detection
Object detection and recognition
Image classification and automated medical diagnosis (using X-rays, CT scan, MRI, ultrasound, microscopic and dermoscopic images)
Hybrid clustering models
Evolutionary algorithms and soft computing techniques (e.g. Genetic Algorithm and Evolutionary Programming)
Signal and image processing
Facial expression recognition
Human action recognition
Image/video captioning
Visual question generation and answering
Image reconstruction and synthesis
Feature extraction and selection
Visual perception and learning
Health monitoring and surveillance
Machine learning, deep learning, and transfer learning for computer vision and medical imaging
Evolving deep architecture generation for computer vision, medical imaging and signal processing problems
Optimal hyper-parameter identification for deep learning, transfer learning, and other classification and regression models
Optimal topology generation for machine learning and ensemble learning models