Photo of Manuel Martinez Torres

Manuel Martinez Torres

Research Assistant at the:
Computer Vision for Human-Computer Interaction Lab (CV:HCI)
Karlsruhe Institute of Technology, Institute for Anthropomatics and Robotics (IAR)

Publications: Google Scholar Profile

22.12.2016: Paper accepted as oral at the Data Compression Conference (DCC 2017)!

Marlin: A High Throughput Variable-to-Fixed Codec using Plurally Parsable Dictionaries

20.12.2016: Paper accepted at the Winter Conference on Applications of Computer Vision (WACV 2017)!

Breath Rate Monitoring during Sleep from a Depth Camera under Real-life Conditions


Robots have always been a passionate source of inspiration for me, and I thrive for the day when sharing the streets with androids is as mainstream as computers are now. My plan is to help this happen as soon as possible.

One of the biggest unsolved challenges Robots must face when dealing with humans is perceiving our unstructured world. This has been my main research focus since 2008.

Since 2010, I am also developing computer assisted technology for the elderly and the visually impaired. In both cases, our goal is to improve their quality of live by helping them being independent and self-sufficient.


Studied both Communications Enginnering and Computer Science at Universitat Politécnica de Catalunya. Obtained a Masters on Robotics at the IRI institute while visiting Carnegie Mellon University and collaborating with Intel Research Pittsburgh under the advising of Antoni Grau, Takeo Kanade, and Siddharta Srinivasa respectively.

Currently, I am pursuing my PhD at Karlsruhe Institute of Technology, being advised by Rainer Stiefelhagen.



I'm an avid programmer, having participated in several competitions since I was a kid. For this, I must sincerely thank professor Salvador Roura. He trained me for the ACM programming contests, and basically cleaned, focused and exponentially enhanced my skills. I particularly enjoy non-conventional pogramming contests, like the 24 hours competitions of Challenge24, and IEEExtreme, and the Obfuscated C Code Contest. Now I'm organizing thematic programming competitions for lectures and conventions.

Object Recognition

During my internship at Carnegie Mellon, I stumbled upon the Object Recognition problem from Alvaro Collet and Siddharta Srinivasa. Super-nice and incredible talented fellows. Alvaro had just improved the object recognition algorithm from David Lowe by doing what now is obvious but nobody had tried before: splitting the keypoints in clusters. Divide and Rule. That was the origin of MOPED, which made object recognition practical in real world applications. I ported the code from matlab to c++ and applied some optimizations on my own in order to make it real time on a single computer.

Automated Monitoring of Sleep

This is important. Bed monitoring. The VIPSAFE project was about monitoring patients in bed to detect dangerous behavior. Thanks to the very protective german laws about privacy and data security, it took some time to be allowed to capture real data. We used this time to create a customized monitoring device (left). Then, we later realized: beds are crucial. Beds are the minimum common denominator: everybody sleeps. We can monitor hospital patients, providing objective and comparable numbers for behavioral aspects such as agitaton and stress. We can monitor elderly care homes, and visualize the progress of patients through this indicators over very large timeframes. Or we can monitor our elderly familiars, to see if they went to bed, or signal an alarm in case a problem is detected. All combinations are possible, and the bed is the common denominator. Beds are the anchor that adds structure to out lives, easing the task of developing computer algorithms that take care of us. I am motivated.

Assistive Technology for the Blind

Well, it didn't take long to realize that computer vision researchers match well with assistive technology for the visually impaired. In particular, people with specialization in computer vision in unstructured environments: human environments. Therefore, it is great that Professor Stiefelhagen drives our Computer Vision for Human-Computer Interaction Lab, and is also the head of the Center of Studies for the Visually Impaired.

Cameras and Sensors

I am pursuing a personal endeavor developing customized camera systems. My focus is on multicamera, smart, and/or embedded systems. On top of that, I have now a low-level knowledge of consumer-level depth cameras, which I modeled in precision in two different papers.

Main Research Projects

SPHERE 2014-2016 Schlafüberwachung im Pflege- und Heimbereich mittels Remotesensorik

SPHERE builds upon the research done in VIPSAFE, extending to Elderly Care the monitoring system we originally designed for Intensive Care Units. I only have nice words to describe our wonderful collaborators, the Evangelische Heimstiftung, and the Thoraxklinikum Heidelberg, which have been extremelly helpfull allowing us to evaluate our systems in real scenarios, and compare our system to the measures obtained by the sleep laboratory Polysomnogram.

More information at: project web page

VIPSAFE 2010-2013 Automated Visual Monitoring for Improving Patient SAFEty

The main goal of the VIPSAFE project is to improve patients’ safety by the means of automated visual monitoring. The project was driven by a German group focused on Intensive Care Units, and a Turkish group focused on Elderly Care.

More information at: project web page


CV:HCI Praktikum

Together with Boris Schauerte and Angela Constantinescu, we started in summer 2011 to offer a practical course about implementing computer vision algorithms. On this course we tackled typical computer vision problems in human environments: reflexions, illumination problems, unstructured environments, etc. Daniel Koester joined the team in 2013 and is driving the praktikum now: web

CV:HCI Programming Assignments

Within the CV:HCI courses imparted by Prof. Stiefelhagen, I collaborate with the Kinect section, and also providing programming challenges for the students. Current problems deal with skin color recognition, person/non-person classification, and face identification.


Marlin: A high throughput variable-to-fixed codec using plurally parsable dictionaries

M. Martinez, M. Haurilet, R. Stiefelhagen, and J. Serra-Sagristà
2017 Data Compression Conference (DCC) pdf bibtex

Breath rate monitoring during sleep from a depth camera under real-life conditions

M. Martinez and R. Stiefelhagen
2017 Winter Conference on Applications of Computer Vision (WACV) pdf bibtex

Cordless sleep monitoring from a single depth camera

A. Benz, M. Martinez, T. Grimm, and R. Stiefelhagen
2016_German Sleep Society (DGSM) webpage

Sleep position classification from a depth camera using bed aligned maps

T. Grimm, M. Martinez, A. Benz, and R. Stiefelhagen
2016 International Conference on Pattern Recognition (ICPR) pdf bibtex

Relaxed earth mover’s distances for chain-and tree-connected spaces and their use as a loss function in deep learning

M. Martinez, M. Haurilet, Z. Al-Halah, M. Tapaswi and R. Stiefelhagen
2016 ArXiv:1611.07573

The SPHERE project: Sleep monitoring using computer vision

M. Martinez and R. Stiefelhagen
2016 Forum Bildverarbeitung pdf bibtex

A closed-form gradient for the 1D earth mover’s distance for spectral deep learning on biological data

M. Martinez, M. Tapaswi, and R. Stiefelhagen
2016 ICML Workshop on Comp. Biology (CompBio) Best Poster Award pdf bibtex

Action recognition in bed using BAMs for assisted living and elderly care

M. Martinez, L.Rybok, R. Stiefelhagen
2015 IAPR International Conference on Machine Vision Applications (MVA) pdf bibtex

Cognitive evaluation of haptic and audio feedback in short range navigation tasks

M. Martinez, A. Constantinescu, B. Schauerte, D. Koester, R. Stiefelhagen
2014 International Conference on Computers Helping People with Special Needs (ICCHP) pdf bibtex

Kinect Unbiased

M. Martinez, R. Stiefelhagen
2014 International Conference on Image Processing (ICIP) pdf bibtex

AKTIV: Multimodal interaction system to engage patients with dementia

F. Putze, M. Tapaswi, M. Martinez, D. Telaar, D. Heger, S. Sarfraz, T. Schultz, R. Stiefelhagen
2014 Chapter in Technische Unterstützung für Menschen mit Demenz, KIT Scientific Publishing webpage

Way to go! detecting open areas ahead of a walking person

B. Schauerte, D. Koester, M. Martinez, R. Stiefelhagen
2014 ECCV Workshop on Assistive Computer Vision and Robotics (ACVR) pdf bibtex

“BAM!” depth-based body analysis in critical care

M. Martinez, B.Schauerte, R. Stiefelhagen
2013 International Conference on Computer Analysis of Images and Patterns (CAIP) pdf bibtex

Automated multi-camera system for long term behavioral monitoring in intensive care units

M. Martinez, R. Stiefelhagen
2013 IAPR International Conference on Machine Vision Applications (MVA) pdf bibtex

Kinect unleashed: Getting control over high resolution depth maps

M. Martinez, R. Stiefelhagen
2013 IAPR International Conference on Machine Vision Applications (MVA) pdf bibtex

XEOTV: 1080P underwater geotagger video with deferred filtering

M. D. L. Reguera, M. Martinez, and X. Fernandez Hermida
2013 Instrumentation Viewpoint / Martech 2013 webpage

Breath rate monitoring during sleep using near-ir imagery and PCA

M. Martinez, R. Stiefelhagen
2012 International Conference on Pattern Recognition (ICPR) pdf bibtex

An assistive vision system for the blind that helps find lost things

B. Schauerte, M. Martinez, A. Constantinescu, and R. Stiefelhagen
2012 Int. Conference on Computers for Handicapped Persons (ICCHP) pdf bibtex

The MOPED framework: Object recognition and pose estimation for manipulation

A. Collet, M. Martinez, S. Srinivasa
2011 International Journal of Robotics Research (IJRR) webpage

MOPED: A Scalable and Low Latency Object Recognition and Pose Estimation System

M. Martinez, A. Collet, S. Srinivasa
2010 International Conference on Robotics and Automation (ICRA) webpage