Teo de Campos Publications Students Projects Datasets Teaching Demos CV I am a Senior Applied Scientist at Microsoft Applied Sciences Group. Prior to Microsoft, I have worked as a Computer Vision / Deep Learning Scientist at Vicon Motion Systems (a member of the Oxford Metrics Group, from 2022 to 2023) and I was also a professor at the University of Brasilia (UnB) from 2016 to 2024. I am also linked with the CVSSP/University of Surrey, as a "visiting researcher", where I had worked as a senior research fellow from 2009 to 2016. From March 2013 to April 2014 I'd also worked in the Machine Learning group at the University of Sheffield. Previously, I have worked in the research laboratories of Xerox (now Naver), Microsoft and Sharp. I completed my doctorate (DPhil) in 2006 at the University of Oxford and my award winning MPhil thesis at the University of Sao Paulo, in 2001. My research interests include image and video generation, computer vision, natural language processing, transfer learning, anomaly detection, tracking and 3D understanding for spatial audio and other machine learning applications.

Publications

Google scholar is probably the most up to date page for my list of publications, but it may not point to freely available PDF files or my papers.
The links below offer alternatives, which may include links to demonstrations, datasets, etc.

• Google scholar
• Here is a big, ugly and unsorted list of references in BibTeX format.
• My thesis, old demos and publications from my page at Oxford
• DBLP
• ORCID
• arXiv
• Lattes (Brazil)
• Surrey University library
• My papers at the old CVSSP library
• LinkedIn
• ResearchGate
• MS academic
• My old page at Surrey

• Patricia Medyna Drumond, PhD student (2019-present) on text mining.
• Lindeberg Leite, PhD student (2020-present) on epitope prediction.
• Aloisio Dourado Neto, PhD student (2018-present) on 3D semantic scene completion.
• Luciano Cejnog, PhD student (co-supervision since 2017) on 3D hand tracking for occupational therapy.
• Frederico Guth, MSc student (2019-present) on transfer learning.
• Pedro Henrique Luz de Araujo, MSc student (2019-present) on document classification.
• Tiago de Carvalho Gallo Pereira, MSc student (2019-present) and final year UG student (2018-2019) on person re-identification.
• Raphael Soares, final year UG student (2019) on image compression.
• Guilherme Lopes, final year UG student (2019) on semantic scene completion.
• André B. S. Guedes was an undergrad student who has worked with me from September 2017 to July 2018 on these applications of Deep CNN: gesture recognition (see his TCC1 report, in Portuguese), concept-based image retrieval and on 3D semantic scene completion from and RGB image and its depth map.
• Moacir Ponti, is a Professor Assistente at the Universidade de Sao Paulo who took a sabbatical at Surrey (Feb 2016 to Jan 2017), working with Josef Kittler, myself and other researchers at the CVSSP. We have collaborated in a project on classifier incongruence, in which his undergrad student Mateus Riva was also involved (as a visiting student at Surrey).
• Sam Fowler, is a PhD student working on indoors scene analysis who started in October 2015, supervised by Adrian Hilton and co-supervised by me until June 2016. After my move to Brasilia, Hansung Kim became his co-supervisor.
• Nazli FarajiDavar was a PhD student who worked on Transfer Learning (Unsupervised Domain Adaptation). She was supervised by me and co-supervised by Josef Kittler from September 2010 to February 2015.
• Estephan Dazzi was a visiting student who worked with me from May 2014 to April 2015, his PhD at USP was supervised by Roberto M. Cesar-Jr. We still collaborate in a project on keygraph matching.
• Mariana Nunes de Oliveira was a Summer placement student from UnB who worked with me from June-September 2014.
• Dalia Coppi was a visiting student in 2012-2013, I was a co-supervisor in her visit, which was mainly supervised by Josef Kittler. Her PhD at UNIMORE was supervised by Rita Cucchiara.
• Iacer Calixto's work in our V&L project (2012) was co-supervised by me. Lucia Specia was his main supervisor.
• Marcelo Mergulhao Russi was my intern at Uana Tech in 2011.
• Julian McAuley was my intern at Xerox in 2009. His PhD at NICTA was supervised by Tiberio Caetano.

Projects

Here is a list of projects, showing the period in which I have been involved (not necessarily the duration of the project as a whole):

• I am the Principal Investigator of project KnEDLe - Knowledge Extraction from Documents of LEgal content (2020-2022)
• Follow this link to see the projects of UnBVision, my group at GitLab, which include the projects below (among others):
• Semantic Scene Completion from 360 cameras (2019-present)
• EdgeNet - 3D Semantic Segmentation using depth and 2D edges.
• TF-SSCnet, Aloisio's implementation of SSCNet (which was originally written in Caffe) using TensorFlow. This implementation has a much smaller memory footprint than the original code and it also separates pre-processing steps from the neural net pipeline, making training a lot faster.
• Domain Adaptation for person re-identification (2018-present).
• SkinNet - our deep CNN for skin detection
• Victor Project: an AI system for Brazil's supreme court (2018-2019)
• Hand tracking for occupational therapy (2014-present)
  • Link to the grant for collaboration between Surrey and USP, sponsored by FAPESP.
  • Download a 3D model of my hand (requires an SDL viewer), built by Phil Krejov.
  • Gesture recognition using deep CNNs, by André Guedes.
• Concept-based image retrieval using Deep Convolutional Neural Networks (2016-2017)
• S3A (2014-2019)
  • Semantic scene completion in 3D. Related papers and code:
    • EdgeNet360: Semantic Scene Completion from a Single 360 degree Image and Depth Map
    • EdgeNet: Semantic Scene Completion from RGB-D images
    • Semantic Scene Completion Combining Colour and Depth: preliminary experiments
  • Indoors scene analysis
  • RGBD semantic segmentation using Deep CNNs [demo] and 3D scene completion from a depth image
  • Talker tracking
• Hand tracking for occupational therapy (2014-2017)
• Mitosis detection (as part of my contribution to the SYNERGY project while I was in Sheffield, 2013-2014)
• Visual Media Research Platform Grant (2013-2014)
• Novel (sub-)class detection (2012-2013)
• ACASVA (2009-2013) and Transfer Learning - source code available in this link (2010-2014)
• PASCAL2 Network of Excellence (2009-2013)
• Vision4Translation - Vision and Language (2012)
• Uana Tech - textual and visual plagiarism detection (2010-2014)
• PinView (2008-2009)
• Character recognition in natural scenes (2007)
• Dual view displays (2005-2007)
• Human hand tracking using kinematic and appearance models (2001-2005)
• Feature selection for face recognition (1999-2001)

Relevant activities I've been involved with:

• Sibgrapi 2020 tutorials chair. During the event, I hosted the tutorial of Cordeiro and Carneiro, on deep learning with noisy labels [video | code, paper and datasets].
• I was one of the Area Chairs of WACV 2017.
• From 2009 to 2013, I'd served as Surrey's local manager for the PASCAL2 Network of Excellence, an European Union Network of Excellence which involved 1072 researchers.
• I have helped with the organisation of BMVC 2012 and I've chaire the students' workshop.
• I've served as one of the area chairs of VISAPP 2012.
• I was a member of the V&L net. As part of it, I'm co-leading a pump-priming project on the use of Vision as context for Machine Translation.
• I have co-chaired the Computer Vision sessions at the EURO 2012 conference.
• I regularly review papers for CVPR, ICCV, ECCV, NIPS, BMVC, PAMI, PR and PRL, among other top vision conferences and journals.
• I've ridden a unicycle for the benefit of research: Imre et al 3DIMPVT 2011, videos.
• I've been trying to learn how to teach. Here is what happens when I try to teach something without much preparation.

Datasets

• Victor dataset
• LeNER-Br: a Dataset for Named Entity Recognition in Brazilian Legal Text
• Audio-visual tracking
• Vision4Translation (VLnet) - Using images as context for machine translation
• ACASVA actions - tennis and badminton actions
• Chars74K - characters used in English and Kannada

Teaching

This section is to be constructed, but you can follow this link to see a list of courses in the field of Artificial Intelligence and Machine Learning that are offered at UnB.

Demos and Media Appearances

	Intrudocing myself to Computer Vision students (in Portuguese, August 2020).
	I've featured in a local TV channel (UnB TV), discussing Deep Learning with my colleague Celia Ghedini Ralha (in Portuguese, May 2017).
	I've contributed to a NewScientist article about visual sourveillance, by Tim Revell (January 2017).

Concept-based image retrieval using a Deep Convolutional Neural Network

• Online demo
• Source code

Head/Face Tracking

3D Trackers from my PhD

3D articulated object tracking (demos from my thesis) 3D head tracking to control a wearable robot (work lead by Walterio Mayol and Ben Tordoff, supervised by David Murray)

ACASVA project videos

In the video below, I present a short summary of some of the main aspects of this project:

The video below demonstrates tracking, court detection, homography computation and event detection:

Here is another video, showing the annotation tool that I have developed:

Images as sets of locally weighted features

Follow this link for a talk I presented at Microsoft Research in Cambridge, 2010.