Tony Joseph

Ensemble Squared: A Meta AutoML System

This work presents Ensemble Squared, a "meta" AutoML system that ensembles at the level of AutoML systems. Ensemble Squared exploits the diversity of existing, competing AutoML systems by ensembling the top-performing models simultaneously generated by a set of them. Our work shows that diversity in AutoML systems is sufficient to justify ensembling at the AutoML system level.
arXiv 2020 (submitted to DARPA-D3M research under AutoML Infrastructure)

Jason Yoo, Tony Joseph, Dylan Yung, S. Ali Nasseri, Frank Wood

PDF Project Page bibtex

                              @inproceedings{yoo2019ensemble,
                                    title={Ensemble Squared: A Meta AutoML System},
                                    author={Jason, Yoo and Joseph, Tony and Yung, Dylan and Nasseri, Ali and Wood, Frank},
                                    journal={arXiv preprint and D3M Papers},
                                    year={2020},
                              }

EdgeConnect: Generative Image Inpainting with Adversarial Edge Learning

This work proposes a two stage adversarial model EdgeConnect that comprises of an edge generator followed by an image completion network. The edge generator hallucinates edges of the missing region (both regular and irregular) of the image, and the image completion network fills in the missing regions using hallucinated edges as a priori.
arXiv 2019 (accepted at ICCV-AIM 2019)

Kamyar Nazeri, Eric Ng, Tony Joseph, Faisal Z. Qureshi, Mehran Ebrahimi

PDF Project Page Code bibtex

                                @inproceedings{nazeri2019edgeconnect,
                                      title={EdgeConnect: Generative Image Inpainting with Adversarial Edge Learning},
                                      author={Nazeri, Kamyar and Ng, Eric and Joseph, Tony and Qureshi, Faisal and Ebrahimi, Mehran},
                                      journal={arXiv preprint},
                                      year={2019},
                                }

Joint Spatial and Layer Attention for Convolutional Networks

In this work, we propose a novel approach that learns to sequentially attend to different Convolutional Neural Networks (CNN) layers (i.e., “what” feature abstraction to attend to) and different spatial locations of the selected feature map (i.e., “where”) to perform the task at hand.
arXiv 2019 (accepted at BMVC 2019, Spotlight)

Tony Joseph, Konstantinos G. Derpanis, Faisal Z. Qureshi

PDF Project Page bibtex

                                    @inproceedings{joseph2019JSLAN,
                                          title={Joint Spatial and Layer Attention for Convolutional Networks},
                                          author={Joseph, Tony and Derpanis, Konstantinos and Qureshi, Faisal},
                                          journal={Conference on the British Machine Vision Association (BMVC), 2019},
                                          year={2019},
                                    }

MSc. Thesis

Joint Spatial and Layer Attention for Convolutional Networks
Abstract: In this work, we propose a novel approach that learns to sequentially attend to different Convolutional Neural Networks (CNN) layers (i.e., “what” feature abstraction to attend to) and different spatial locations of the selected feature map (i.e., "where") to perform the task at hand. Specifically, at each Recurrent Neural Network (RNN) step, both a CNN layer and localized spatial region within it are selected for further processing. We demonstrate the effectiveness of this approach on two computer vision tasks: (i) image-based six degree of freedom camera pose regression and (ii) indoor scene classification.
e-scholar@UOIT (Electronic Theses and Dissertations, 2019)

Tony Joseph

PDF bibtex

                                @inproceedings{joseph2019JSLAN,
                                      title={Joint Spatial and Layer Attention for Convolutional Networks},
                                      author={Joseph, Tony},
                                      journal={Electronic Theses and Dissertations (Public), 2019},
                                      year={2019},
                                }

Tony Joseph

AI Engineer

🧠🤖🔧

Interests

Education

Latest News

📚 Publications

Ensemble Squared: A Meta AutoML System

EdgeConnect: Generative Image Inpainting with Adversarial Edge Learning

Joint Spatial and Layer Attention for Convolutional Networks

MSc. Thesis

🧗‍♂️ Career Journey

About Me 💬