Ariyan Zarei

MegaStitch: Robust Large Scale Image Stitching

A Zarei, E Gonzalez, N Merchant, D Pauli, E Lyons, K Barnard
Published in IEEE TGRS.

NowCasting-nets: Representation Learning to Mitigate Latency Gap of Satellite Precipitation Products using Convolutional and Recurrent Neural Networks

MR Ehsani, A Zarei, HV Gupta, K Barnard, E Lyons, A Behrangi
Published in IEEE TGRS.

Improve captcha's security using gaussian blur filter

A Zarei
Published in Airccse Signal & Image Processing Journal Vol. 5, no. 5 (2014): p 35

A Feature Vector for Optical Character Recognition

A Zarei, AY Shooshtari
Published in the proceedings of the 2018 International Conference on Information Science and System

The Effect of Applying Gaussian Blur Filter on CAPTCHA’s Security

A Zarei
Published in Airccse Signal & Image Processing Journal

Computer Controlling System Using Text Messages (Technical Report)

A Zarei
International Conference On Computers, Information Technologies and Digital Medias (CITADIM 2013)

AG2PI Workshop Series

Introduction to Scientific Computing

We presented a three session workshop series in AG2PI on using computer science tools for scientific computing. You can find the slides and some more information about the webinar here.

AG2PI Workshop

Hands-On Machine Learning with Agricultural Applications

In this talk, we introduced basics of machine learning and discussed using machine learning methods to segment and classify images of plants. Find out more by watching the webinar on Youtube. You can find the slides and some more information about the webinar here.

Cyverse Webinar

Managing the Machine Learning Lifecycle with MLFlow: A Demonstration Using PhytoOracle

We presented on Cyverse Webinar the MLflow and its use cases. In this talk, we introduced MLflow as an open source Machine Learning Life Cycle Managament tool and illustrated a use case of it on a plant-science-based semantic segmentation project. Find out more by watching the webinar on Youtube. You can find the slides and some more information about the webinar here.

IVILAB Colloquium

Literature Review on Image Stitching and Alignment

I presented a brief literature review on image stitching and alignment during IVILAB colloquium. I discussed the basics of image stitching and alignment and thereafter reviewed 4 papers with interesting innovations in multiple image alignment and stitching. You can find the slides here.

Agricultural Genome to Phenome Initiatives (AG2PI)

PhytoOracle: A Case Study in Automating Phenotyping

I presented the Image Orthomosaicing method that we have developed in PhytoOracle, i.e. the MegaStitch, in the AG2PI field day seminar. AG2PI or Agricultural Genome to Phenome Initiatives hosts monthly seminars on the correlations between genetipic and phenotipic information. You can find the recorded version of the presentation on Youtube and the AG2PI page.

IVILAB Colloquium / CSC 696H AI Seminar

Different approaches in estimating and calculating the jacobian matrix for nonlinear least squares problem

I presented a study on the effects of using different methods for estimating and calculating the Jacobian matrix of a nonlinear least squares equation for minimization of projection error in a Bundle Adjustment problem. You can find the slides here.

TRIPODS

PhytoOracle: Transforming the Way We See Plants

I presented the MegaStitch method in the TRIPODS Seminar. I discussed the details of our proposed methods for stitching the images in large scale. I illustrated the efficiency and accuracy of MegaStitch in geo-correcting agricultral images captured by drones and ground-based platforms like the Gantry. You can find the slides here.

IVILAB Colloquium / CSC 696H AI Seminar

Bundle Adjustment - A modern Synthesis

I presented the paper by Triggs, Bill et. al. on Bundle Adjustment in IVILAB Colloquium. I discussed different formulations of the bundle adjustment problem and different methods for estimating the parameters of those bundle adjustment problems. You can find the slides here.

Cyverse Webinar

PhytoOracle: Leveraging CyVerse, Containers, and Open-Source Tools for Phenomic Data Processing at Scale

I presented the geo-correction and image stitching as part of our PhytoOracle pipeline presentation in the Cyverse webinar. We discussed different parts of the pipeline and I specifically focused on explaining the stitching method I develped. Find out more here and watch the webinar on youtube here.

IVILAB Colloquium / CSC 696H AI Seminar

MGRAPH Method and Using Non-Linear Least Squares in Large Scale Image Stitching

I presented one of my research projects in IVILAB Colloquium. I talked about large scale image stitching in high level and discussed the MGRAPH method in detail. You can find the slide at here.

IVILAB Colloquium / CSC 696H AI Seminar

Second Order Methods and Conjugate Gradient Method in Optimization

In this presentation, I discussed Newton's Method for optimization as a second order method and I also discussed the Conjugate Direction and Conjugate Gradient Methods as a way to improve the Steepest Descent and Gradient Descent Method. You can find the slide at here.

IVILAB Colloquium

Layer-wise Relevance Propagation in Neural Networks to have more interpretable Machine Learning models

I presented Layer-wise Relevance Propagation technique in Neural Networks during the weekly IVILAB Colloquium. In this technique which is used to provide explainability for Neural Networks, the prediction is propagated backward through the network until the input layer and the regions that the network is using to produce the prediction (positively and negatively) is highlighted using the propagated relevances. You can download the slide here.

IVILAB Colloquium

Structured SVMs and Conditional Random Fields for Semantic Segmentation

I presented a paper about Structured SVMs and using them for semantic segmentation during the weekly IVILAB Colloquium. In this paper, they have explained the mathematical background behind a special type of Support Vector Machines for classifying sequences of data. They also discussed various techniques for training their model. Interestingly, they incorporated Deep Neural Representations into their model for the sake of feature extraction. You can find out more about this technique and its relevance to my research in the slides.