Javier Cruz-Mota was born in Barcelona, Spain, in 1982. He holds a Mathematics degree (2005) and a Telecommunications Engineering degree (2006), both from the Universitat Politècnica de Catalunya (UPC), where he was admitted in a special program for pursuing both degrees simultaneously. In 2006, he obtained a grant for developing his Master Thesis of Telecommunications Engineering at the Signal Processing Laboratory 5 (LTS5), at the École Polytechnique Fédérale de Lausanne (EPFL) in Lausanne, Switzerland.
In January 2007, Javier started his PhD studies in the Electrical Engineering program at EPFL, under the supervision of Prof. Michel Bierlaire and Prof. Jean-Philippe Thiran. At the same time, he worked as Research and Teaching Assistant at the Transport and Mobility Laboratory (TRANSP-OR) at EPFL. His main research topics during his PhD studies were in the fields of image/signal processing and computer vision. In particular, he worked on visual tracking, mathematical modelling and simulation, omnidirectional vision and behavioural information in image analysis. In August 2011, he successfully defended his PhD Thesis, entitled "Model-based Behavioural Tracking and Scale Invariant Features in Omnidirectional Matching" (see Publications), and in September 2011 he received his PhD degree.
After a short period working for a medical devices company, Javier joined an innovation Team at Cisco, the Self Learning Networks (SLN) Team. There, Javier developed between 2013 and 2017 Analytics, Machine Learning and Deep Learning algortihms for extracting insights and predicting trends in networking data, with applications in the fields of IoT, security for IoT and security for Enterprise networks. The first product fully developed by the SLN Team was the Cisco Stealthwatch Learning Network License, where Javier contributed from the conception until the final implementation phases. Two years ago, Javier joined the Kairos Engineering Team, which is the Team behind Cisco AI Network Analytics, a cloud-based AI-driven service for detecting issues, extracting trends and generating insights from data of large wireless networks. In this Team, Javier leads the ML/Analytics Group, which is in charge of designing and developing the data storage, processing and learning pipelines.
Javier is author/co-author of several international journal and conference papers (see Publications for details) and of more than 40 patents (32 already issued) in the fields of machine learning and networking (see Patents for further details). Furthermore, he acts from time to time as reviewer for some international journals such as Computer Vision and Image Understanding, IEEE Transactions on Circuits and Systems for Video Technology, IEEE Signal Processing Letters, Signal, Image and Video Processing, IEEE Transactions on Image Processing and SPIE Optical Engineering.
Outside academic and professional aspects, Javier is a meteorology and astronomy lover, and enjoys running, skiing and spending time with his family.
- Network attack detection using combined probabilities, J. Cruz Mota, A. Di Pietro and J.-P. Vasseur. [Full text here]
- Traffic segregation in DDoS attack architecture, J.-P. Vasseur, A. Di Pietro and J. Cruz Mota. [Full text here]
- Applying a mitigation specific attack detector using machine learning, J. Cruz Mota, A. Di Pietro and J.-P. Vasseur. [Full text here]
- Distributed approach for feature modeling using principal component analysis, J. Cruz Mota, J.-P. Vasseur and A. Di Pietro. [Full text here]
- Hierarchical event detection in a computer network, J.-P. Vasseur, J. Cruz Mota and A. Di Pietro. [Full text here]
- Learning model selection in a distributed network, J.-P. Vasseur, A. Di Pietro and J. Cruz Mota. [Full text here]
- Anomaly detection in a computer network, J.-P. Vasseur, J. Cruz Mota and A. Di Pietro. [Full text here]
- Control loop control using broadcast channel to communicate with a node under attack, J.-P. Vasseur, J. Cruz Mota, A. Di Pietro and J.W. Hui. [Full text here]
- Attack mitigation using learning machines, J.-P. Vasseur, J. Cruz Mota, A. Di Pietro and J.W. Hui. [Full text here]
- Quarantine-based mitigation of effects of a local DoS attack, J. Cruz Mota, J.-P. Vasseur and A. Di Pietro. [Full text here]
- Using learning machine-based prediction in multi-hopping networks, A. Di Pietro, J.-P. Vasseur and J. Cruz Mota. [Full text here]
- Distributed voting mechanism for attack detection, J.-P. Vasseur, A. Di Pietro and J. Cruz Mota. [Full text here]
For a list of all my co-authored patents (issued and under review) visit this link
Some small pieces of code tat I developed mostly during my PhD studies.
Scale Invariant Feature Transform on the Sphere (SIFTS)
The SIFTS package is an implementation in MatLab of the algorithm introduced in "Scale Invariant Feature Transform on the Sphere: Theory and Applications" (see Publications). This algorithm performs a scale invariant feature transform on omnidirectional images.
A COnvenient Bunch of Instructions (COBI)
COBI is a small multithread and multiplatform interface where you can easily (I hope!) plug a video processing algorithm to test its performance, read input videos, write output videos... I developed this interface during my PhD studies for testing my people tracking algorithms. COBI uses Qt for the interface and the thread management, FFMPEG for the video reading and writing, and OpenCV for the frame processing. See the README file for further details.
A simple and educational program to perform a Principal Component Analysis (PCA) of data.
Web MOSix MONitor (WMOSMON)
A web-based MOSIX Monitor written in PHP. This web interface was developped in 2002 and has not been updated since then, so do not be surprised if it does not work :-)
Some demo videos of the tracking algorithms that I developed during my PhD.
Example of face tracking using IVT, ITWVT/M (see Chapter 3 of my PhD Thesis in Publications) and TLD:
MOTOH Pedestrian Tracker
Examples of pedestrian tracking in the presence of total occlusions. Results obtained using IVT, MOTOH (see Chapter 4 of my PhD Thesis in Publications) and TLD.
Real and unpredictable occlusion: