Thomio Watanabe
PhD in Computer Science
Salvador, Brazil

I'm a software engineer with a PhD in computer science and experience in robotics and perception systems development for autonomous cars and autonomous underwater vehicles.

Saipem FlatFish (from 2020). Industrialization process of the FlatFish autonomous underwater vehicle. Main activities comprise the perception system development for pipeline and riser inspection. Also assisting the simulation and underwater communication.

• Winner of the 2023 OTC Spotlight on New Technology Award. [link]
• Winner of the 2023 OMC Energy of Tomorrow Award, with the paper: Autonomous deepwater inspections of risers via subsea drones and A.I: based data fusion tools [link]

Carina 2 (2016-2021). PhD project in autonomous cars at USP-ICMC, Brazil. Main topic: deep learning for obstacles detection in images.

• Image classification in frequency domain with 2SReLU: a second harmonics superposition activation function. [paper]
• Instance Segmentation as Image Segmentation Annotation. [paper]



• Verisimilar Percept Sequences Tests for Autonomous Driving Intelligent Agent Assessment. [paper]
• Distance to Center of Mass Encoding for Instance Segmentation. [paper]

FlatFish project (2014 - 2015). 2 years project at SENAI/CIMATEC - Brazilian Institute of Robotics. Autonomous Underwater Robot development for oil & gas structures inspection. Main activities: integration between robotics framework and simulator (rock-gazebo); underwater environment modeling and simulation; 3D model reconstruction of underwater assets.

• FlatFish - a compact subsea-resident inspection AUV. [paper]
• The Rock-Gazebo Integration and a Real-Time AUV Simulation. [paper]
• 3D model reconstruction of underwater assets. [code example]

Publications

• T. Watanabe and D. F. Wolf, Image classification in frequency domain with 2SReLU: a second harmonics superposition activation function in Applied Soft Computing, 2021. [arxiv, bibtex]
  @article{watanabe2021srelu, title = {Image classification in frequency domain with 2SReLU: A second harmonics superposition activation function}, journal = {Applied Soft Computing}, volume = {112}, pages = {107851}, year = {2021}, issn = {1568-4946}, doi = {https://doi.org/10.1016/j.asoc.2021.107851}, url = {https://www.sciencedirect.com/science/article/pii/S1568494621007730}, author = {Thomio Watanabe and Denis F. Wolf}, keywords = {Image classification, Artificial neural networks, Activation function, Frequency domain} }

• T. Watanabe and D. F. Wolf, Instance Segmentation as Image Segmentation Annotation in 30th IEEE Intelligent Vehicles Symposium, Paris, 2019. [arxiv, poster, bibtex]
  @inproceedings{watanabe2019instance, title={Instance Segmentation as Image Segmentation Annotation}, author={Thomio Watanabe and Denis F. Wolf}, booktitle={30th IEEE Intelligent Vehicles Symposium}, address={Paris}, year={2019} }

• T. Watanabe and D. Wolf, Distance to Center of Mass Encoding for Instance Segmentation in The 21st IEEE International Conference on Intelligent Transportation Systems (ITSC), Maui-Hawaii, 2018. [arxiv, presentation, code, bibtex]
  @inproceedings{watanabe2018dcme, title={Distance to Center of Mass Encoding for Instance Segmentation}, author={Thomio Watanabe and Denis Wolf}, booktitle={The 21st IEEE International Conference on Intelligent Transportation Systems (ITSC)}, address={Hawaii}, year={2018} }

• T. Watanabe and D. Wolf, Verisimilar Percept Sequences Tests for Autonomous Driving Intelligent Agent Assessment in 15th Latin American Robotics Symposium and 6th Brazilian Robotics Symposium (LARS-SBR), João Pessoa, 2018. [arxiv, presentation, bibtex]
  @inproceedings{watanabe2018verisimilar, title={Verisimilar Percept Sequences Tests for Autonomous Driving Intelligent Agent Assessment}, author={Thomio Watanabe and Denis Wolf}, booktitle={15th Latin American Robotics Symposium and 6th Brazilian Robotics Symposium (LARS-SBR)}, address={João Pessoa}, year={2018} }

• T. Watanabe et al., The Rock-Gazebo Integration and a Real-Time AUV Simulation in 12th Latin American Robotics Symposium and 2015 3rd Brazilian Symposium on Robotics (LARS-SBR), Uberlandia, 2015, pp. 132-138. doi: 10.1109/LARS-SBR.2015.15. [pre-print, code, bibtex]
  @inproceedings{watanabe2015rock, title={The Rock-Gazebo Integration and a Real-Time AUV Simulation}, author={Thomio Watanabe and Gustavo Neves and Rômulo Cerqueira and Tiago Trocoli and Marco Reis and Sylvain Joyeux and Jan Albiez}, booktitle={12th Latin American Robotics Symposium and 2015 3rd Brazilian Symposium on Robotics (LARS-SBR)}, address={Uberlandia}, year={2015} }

• J. Albiez et al., FlatFish - a compact subsea-resident inspection AUV in OCEANS 2015 - MTS/IEEE Washington, Washington, DC, 2015, pp. 1-8. [bibtex]
  @inproceedings{albiez2015flatfish, title={FlatFish - a compact subsea-resident inspection AUV}, author={Jan Albiez and Sylvain Joyeux and Christopher Gaudig and Jens Hilljegerdes and Sven Kroffke and Christian Schoo and Sascha Arnold and Geovane Mimoso and Pedro Alcantara and Rafael Saback and Joao Britto and Diego Cesar and Gustavo Neves and Thomio Watanabe and Patrick Merz Paranhos and Marco Reis and Frank Kirchner}, booktitle={OCEANS 2015 - MTS/IEEE Washington}, address={Washington}, year={2015} }


Research interests

• Computer Vision & Perception
• Machine Learning & Deep Learning
• Robotics Frameworks & Simulation
• Embedded Systems

Personal projects

• DeepLibs. Website with explanations and interesting facts about deep learning applied in computer vision. The general idea is to explore deep learning concepts and provide code examples in python.
• RL Gazebo. Gazebo plugin to simulate Deep Reinforcement Learning (DQN) in mobile robots.
• Custom FFT. Fast-Fourier Transform implementation in C.
• Git presentation. Hands-on introduction to git.

Awards

• TopCoder - Harvard Medical School lung tumor challenge. The challenge requested participants to detect and delineate lung tumor regions in thoracic computed tomography (CT) scans. Placement: 7th in phase 1 and 1st in phase 2. [phase 1, phase 2, source code 1, source code 2, paper]
• FAPESP PhD scholarship. Three years São Paulo research foundation PhD scholarship.
• Erasmus Mundus scholarship. One year bachelor exchange at EhB/VUB, Belgium. Bachelor thesis development with industrial cooperation.
• BRAFITEC scholarship. One year bachelor exchange at CPE Lyon, France. BAC+4 courses.

Tools

• Programming - C, C++, Python, Git
• Robotics - Linux, ROS/Rock, Gazebo, OpenCV, Microcontrollers
• Machine learning - Caffe, Pytorch, Tensorflow

Languages

• Portuguese - mother tongue.
• English - fluent.
• French - fluent.