JB Boin

Research

Image Descriptor Aggregation for Efficient Retrieval

Ph.D. thesis.

Fall 2013 - Summer 2019.

Defense slides Thesis

Efficient Panorama Database Indexing for Indoor Localization

Research project. Collaboration with Dmytro Bobkov, Eckehard Steinbach and Bernd Girod.

This work was included in the Best Paper Session at CBMI 2019.

Spring - Fall 2018.

CBMI 2019 (paper) CBMI 2019 (presentation)

Code on Github

Recurrent Neural Networks for Person Re-identification Revisited

Research project. Collaboration with André Araujo and Bernd Girod.

This work received the Best Paper Award at MIPR 2019.

Fall 2017 - Winter 2018.

MIPR 2019 (paper) MIPR 2019 (presentation)

Code on Github (partial)

Art++

Art++ is an interactive museum guide using augmented reality on mobile devices. The augmented reality component uses state-of-the-art image retrieval techniques combined with real-time tracking and rendering. The project led to an exhibition at the Cantor Arts Center over the summer 2016.

Art++ was funded by a Magic Grant from the Brown Institute for Media Innovation.

This work got media coverage in the New York Times [1] [2], the Stanford Daily [3] [4], the Palo Alto Weekly [5], The Creators Project [6] and the San Francisco Chronicle [7].

Summer 2014 - Summer 2016.

Art++ is an interactive museum guide using augmented reality on mobile devices. The augmented reality component uses state-of-the-art image retrieval techniques combined with real-time tracking and rendering. The project led to an exhibition at the Cantor Arts Center over the summer 2016.

Art++ was funded by a Magic Grant from the Brown Institute for Media Innovation.

This work got media coverage in the New York Times [1] [2], the Stanford Daily [3] [4], the Palo Alto Weekly [5], The Creators Project [6] and the San Francisco Chronicle [7].

Summer 2014 - Summer 2016.

Watch video

Effective Fisher Vector Aggregation for 3D Object Retrieval

Research project. Collaboration with André Araujo, Lamberto Ballan and Bernd Girod.

Winter - Spring 2016.

ICASSP 2017 (paper) ICASSP 2017 (poster)

Code on Github

Depth Augmented Stereo Panorama

Research project. Collaboration with Jayant Thatte, Haricharan Lakshman, Gordon Wetzstein and Bernd Girod.

This project led to two publications. The conference paper "Depth Augmented Stereo Panorama For Cinematic Virtual Reality With Focus Cues" received the Conference Best Paper Award for Industry at ICIP 2016.

Summer - Fall 2015.

Research project. Collaboration with Jayant Thatte, Haricharan Lakshman, Gordon Wetzstein and Bernd Girod.

This project led to two publications. The conference paper "Depth Augmented Stereo Panorama For Cinematic Virtual Reality With Focus Cues" received the Conference Best Paper Award for Industry at ICIP 2016.

Summer - Fall 2015.

ICME 2016 ICIP 2016

Past projects

                                
                                AI Agent for Light Cycle Racing
                            
AI Agent for Light Cycle Racing in TronCourse project for CS 221 (Artificial Intelligence: Principles and Techniques).
Fall 2015.

                                Report
                            
                                Dissection of a CNN
                            
Tiny ImageNet Challenge - Dissection of a convolutional neural networkCourse project for CS 231N (Convolutional Neural Networks for Visual Recognition).
Spring 2015.

                                Report
                                Poster
                            
                                Feature Tracking on Android
                            
Improving Feature Tracking Using Motion Sensors on AndroidCourse project for CS 231M (Mobile Computer Vision).
Spring 2014.

                                Report
                                Slides
                            
                                Multi-resolution image recognition
                            
Multi-resolution image recognitionResearch project. Collaboration with Roland Angst, David Chen and Bernd Girod.
Summer/Fall 2014.

                                Slides
                            
                                Lost Flock
                            
Lost FlockAndroid game designed as a course project for CS 248 (Interactive Computer Graphics). Collaboration with CJ Cullen, Ranjay Krishna and Jeff Stone.
Spring 2014.

                                Video
                            
                                Plane Rectification on Android
                            
Plane Rectification in Real Time on an Android DeviceResearch project. Also presented as a course project for CS 231A (Computer Vision: From 3D Reconstruction to Recognition).
Fall 2013/Winter 2014.

                                Report
                                Poster
                            
                                Polyphonic Piano Transcription
                            
Polyphonic Piano TranscriptionCourse project for CS 229 (Machine Learning). Collaboration with Mohammad Sadegh Ebrahimi.
Fall 2013.

                                Report
                                Poster
                            
                                Blood Vessel Segmentation
                            
Blood Vessel Segmentation in Retinal Fundus ImagesCourse project for EE 368 (Digital Image Processing). Collaboration with Karianne Bergen and Sahinaz Safari Sanjani.
Spring 2013.

                                Report
                                Poster
                            
                                Camera Forensics
                            
Camera ForensicsCourse project for PSYCH 221 (Applied Vision and Image Systems). Collaboration with Tiffany Jou.
Winter 2013.

                                Report
                                Slides

Personal projects

Wobble

Parallax is a strong cue for depth perception. This project was inspired by gifs such as these ones, where a strong sense of depth is given by alternating between nearby viewpoints. In this demo, I wanted to explore this effect interactively in a virtual 3D world. The focus is automatically locked to the object at the center of the screen, and the camera frequency can be manually adjusted.

This demo was built using Unity.

Demo

3D From Head Tracking

This project was inspired by this video from 2007, which uses head tracking to increase depth perception. Here, we detect the position of the head relative to the screen by using the webcam, assumed to be located above the screen. We then use the clmtrackr library to extract the position of the eyes in real time. This allows us to render the scene as it would be seen from these eye positions if it was in the real 3D world.

This demo was built using three.js and was tested on Chrome and Firefox.

Demo

Teaching

EE 368: Digital Image Processing

Teaching assistant (head teaching assistant in Fall 2016, Winter 2018 and Winter 2019).

Spring 2015, Fall 2016, Winter 2018, Winter 2019.

Website

EE 261: The Fourier Transform

EE 261: The Fourier Transform and its Applications

Head teaching assistant.

Summer 2013.