“Imaging beyond imagination”
That is this year’s ICIP 2018 conference theme. I attended the world’s largest and most comprehensive technical conference focused on image and video processing and computer vision.
Here are my special picks:
Image Representation an Modeling
Reducing Anomaly Detection in Images to Detection in Noise
Smart approach to anomaly detection by removing self-similar content of the image – ready to use for detecting material defects, tumors and others!
Feature Dimensionality Reduction with Graph Embedding and Generalized Hamming Distance
Imagine you have a large database of images with assigned labels or tags. You may want to perform dimensionality reduction taking the labels into account (in order to effectively apply some supervised machine learning methods, to be able to automatically assign labels or find similar images etc).
Experimentally Defined Convolutional Neural Network Architecture Variants for Non-Temporal Real-Time Fire Detection
Large deep convolutional neural networks (CNN) with are not suitable for embedded fire detection devices. Therefore, authors are defining reduced-complexity CNN architecture for real-time full-frame fire detection and in-frame localization.
How Old Do You Look? Inferring Your Age from Your Gaze
Sounds like sci-fi, but it is true for sure – authors can tell your approximate age group by observing your eye movement patterns. An interesting thing to learn is that adults tend to have wide horizontal field of vision.
Knot Magnify Loss for Face Recognition
We have seen many face recognition techniques so far. So, how is this any different? Noah’s Ark Laboratory (Huawei Technologies Co., Ltd) presents Knot Magnify loss in order to magnify the effect of rare hard samples for training.
Linear and Non-linear Filtering
Adaptive transform via quantum signal processing: application to signal and image denoising
Did you know it is possible to use tools from quantum mechanics, such as the Schroedinder equation, in order to construct an adaptive transform suitable for signal and image processing applications, e.g. denoising?
Sensing, Representation and Display
Multilevel Illumination Coding for Fourier Transform Interferometry in Fluorescence Spectroscopy
This very domain-specific talk will take you through the interesting problematic of the acquisition of biological HyperSpectral Volume (HSV) by Fourier Transform Interferometry with non-uniform density sampling. The whole approach is proven with rigorous math!
Deep Learning for Detection and Classification
Deep-Learning-Based Pipe Leak Detection Using Image-Based Leak Features
The presentation is considering deep learning pipe leak detection using trajectory-based image features that reflect the aforementioned characteristics of the acoustic leakage signal.
Segmentation using Deep Learning
Adversarial Domain Adaptation with a Domain Similarity Discriminator for Semantic Segmentation of Urban Areas
In this paper, authors are proposing an adversarial domain adaptation method with a domain similarity discriminator, which consists of a Siamese network, in order to eliminate the domain shift for semantic segmentation of urban areas.
Affine Invariant Image Comparison Under Repetitive Structures
Affine invariant image comparison is a difficult task even without the presence of noise or repetitive structures. Authors will show how to tackle this problem using a local field of image gradient orientation and affine simulations.
Image Registration and Alignment
Scalable Multi-Consistency Feature Matching with Non-Cooperative Games
Don’t forget to check my other picks from the computer vision conference ECCV 2018!