Projects

Project Resources
Saliency-guided Visual Attention Modeling (SVAM) Preprint

Diver Face Recognition for Underwater HRI

We present a deep-learned facial recognition method for underwater robots to identify scuba divers. Specifically, the proposed method is able to recognize divers underwater with faces heavily obscured by scuba masks and breathing apparatus. Our proposed framework is able to learn discriminative features from real-world diver faces through different data augmentation and generation techniques.

Investigator: Jungseok Hong, Sadman Sakib Enan

Preprint

Deep SESR: Simultaneous Enhancement and Super-Resolution 

We introduce and tackle the simultaneous enhancement and super-resolution (SESR) problem for underwater robot vision and provide an efficient solution for near real-time applications:  Deep SESR. We also present UFO-120, the first dataset to facilitate large-scale SESR learning.

Investigator: Md Jahidul Islam, Peigen Luo

Preprint
Code
Data

Semantic Segmentation of Underwater Imagery: Dataset and Benchmark

We introduce the first large-scale dataset for semantic segmentation of underwater imagery (SUIM). We conduct a comprehensive benchmark evaluation of several state-of-the-art semantic segmentation approaches based on standard performance metrics. We also present SUIM-Net, a fully-convolutional deep residual model that balances the tradeoff between performance and computational efficiency.  

Preprint
Code
Data

Fast Underwater Image Enhancement: FUnIE-GAN

We present a fast GAN-based model for real-time underwater image enhancement: FUnIE-GAN. We also present a large-scale dataset named EUVP to facilitate paired and unpaired learning of underwater image enhancement.

Investigator: Md Jahidul Islam, Youya Xia

Paper
Preprint
Code
Data

Underwater Image Super-Resolution using Deep Residual Multipliers: SRDRM

We present a deep residual network-based generative model for single image super-resolution underwater (SISR), which we refer to as SRDRM. We also formulate its adversarial training pipeline, i.e.SRDM-GAN. Additionally, we present USR-248, a large-scale dataset that contains paired instances for supervised training of 2x, 4x, or 8x SISR models. 

Investigator: Md Jahidul Islam, Sadman Sakib Enan, Peigen Luo

Preprint
Code
Data

Balancing Robustness and Efficiency in Deep Diver Detection 

Design and development of a class of autonomous diver-following algorithms that are: a) invariant to color (of divers’ body/wearables), b) invariant to divers’ relative motion and orientation, c) robust to noise and image distortions, and d) reasonably efficient for real-time deployment.

Investigator: Md Jahidul Islam, Michael Fulton

   Paper

Visual Diver Recognition for Underwater Human-Robot Collaboration

This paper presents an approach for autonomous underwater robots to visually detect and identify divers. The proposed approach enables an autonomous underwater robot to detect multiple divers in a visual scene and distinguish between them. 

Investigator: Youya Xia

 

Pre-print

Robot Communication Via Motion

Developing new modalities of interaction for 6DOF robots using motion as the primary robot-to-human communication vector, to be used in challenging environments.

Investigator: Michael Fulton

pre-print

talk

Detection of Marine Debris

Investigating and applying methods for object detection to the detection of marine debris, particularly plastic trash, with the eventual purpose of removal and monitoring. 

Investigator: Michael Fulton, Jungseok Hong

pre-print

SafeDrive 

Investigating robust methods for finding lane markers under severely degraded visibility, for assisted and autonomous driving.

Investigator: Jiawei Mo

  Paper
  Poster

Mixed-Domain Periodic Motion (MDPM) Tracker 

Robust algorithm for an autonomous underwater robot to visually detect and track its companion human diver swimming in arbitrary directions.

Investigator: Md Jahidul Islam

  Abstract
  Paper 
  Poster

RoboChatGest 

A real-time programming and parameter reconfiguration method for autonomous underwater robots using a set of hand-gestures

Investigator: Md Jahidul Islam, Marc Ho

  Paper-1
  Paper-2

Adversarial Image_Colorization

Using Generative Adversarial Networks for automatic image colorization.
Investigator: Cameron Fabbri

  Github

  

Underwater Image Restoration

Using deep learning for restoring underwater images.

Investigator: Cameron Fabbri, Md Jahidul Islam

   Paper

smartTalk: A Learning-based Framework for Natural Human-Robot Dialog

Investigator: Cameron Fabbri

Paper
Poster