[ Patent Deposit – Samsung ] Video Recognition Method Capable of Encoding Concept Temporal and Spatial Relations Using Contextual Information

December 11, 2019

The proposed invention aims to encode contextual information for video analysis and understanding, encoding spatial and temporal relations of objects and the main agent in a scene. The main target application of the invention is the recognition of human activity. The coding of these spatial and temporal relationships can be crucial in distinguishing different categories of human activities and can be important in helping to discriminate between different video categories for video classification, retrieval, categorization, and other video analytics applications...Ler Mais

Making Sense: Robust Approaches to Visual Monitoring

February 15, 2019

Making Sense aims to develop approaches to solving problems related to visual surveillance. It focuses on the aspects of knowledge extraction from surveillance scenarios, including the problems of activity recognition, detection of anomalous events, identity maintenance and optimization of deep neural networks...Ler Mais

SMS: Research and Development of a Intelligent Surveillance System Applied to Oil Platforms

May 1, 2018

This project aims at the creation, development, deployment and experimental validation of a prototype intelligent surveillance system capable of monitoring workers in oil platforms through visual data. This system will be able to issue warnings in cases of inadequate use of protection equipment, obstructions of escape routes and will be capable of biometrically identifying all workers on the oil platform...Ler Mais

V+: Video Analytics Solutions

June 1, 2017

The focus of this project is the development of video analysis solutions for the internet of things (IoT) platform of Maxtrack, the largest tracking and telemetry company in Latin America. Several problems in the surveillance and biometrics domains are addressed in this project, including vehicle recognition and person detection, tracking, and identification...Ler Mais

Keymaker: Activity Recognition Based on Contextual Information

June 1, 2016

Keymaker addressed the research of new algorithms for recognition of human activities via the extraction of information, such as the importance of objects in the execution of the activity and the understanding of the interaction of individuals with objects present in the scene...Ler Mais

HAR-HEALTH: Recognition of Human Activities Associated with Chronic Diseases

February 1, 2016

The objective of this project was the research and development of methods and algorithms capable of automatically recognizing human activities related to chronic diseases (diabetes, hypertension, obesity and aging) from visual information, signals captured by personal mobile device sensors and signals captured by sensors installed in environments...Ler Mais

GigaFrames: Large-Scale Surveillance and Computational Forensics

July 3, 2015

The automatic processing of camera images is essential to assist security agents in crime prevention (surveillance) and search for evidence in cases where crimes have already occurred (Computational Forensics). GigaFrames focuses on the development of computer vision and machine learning techniques to perform environment monitoring and digital forensic analysis from large volumes of data captured through surveillance cameras...Ler Mais

DeepEyes: Visual Computing Solutions and Machine Intelligence for Computer Forensics and Electronic Surveillance

October 1, 2014

This project aimed at the development of algorithmic solutions of visual computation and machine intelligence for problems related to forensic computation, digital security and electronic surveillance. The problems of interest were: detection of falsifications in digital images and videos; detection of clandestine plantations of e.g., Cannabis sativa from remote sensing images, development of human identification techniques from faces, among others...Ler Mais

DET: Efficient Pedestrian Detection Applied to People Observation

February 17, 2014

Automatic pedestrian detection is essential so that the huge amounts of visual data captured by surveillance cameras is reduced to a volume that can be managed by current computing systems. Thus, the activities being performed by agents present in the scene can be analyzed. This project proposed the development of methods for pedestrian detection to reduce computational cost and maintain the accuracy obtained by detectors...Ler Mais

SmartView: Automatic Monitoring Applied to Major Sports Competitions

December 12, 2013

With the purpose of assisting in the monitoring and, consequently, the safety of fans present in sports competitions, this project aimed at using computer vision techniques to automate the detection of conflicts occurring in the bleachers and the identification and location of the individuals involved...Ler Mais

VER+: Robust and Efficient Surveillance Methodologies

March 8, 2013

This project had as its main focus the development and improvement of computer vision techniques to monitor environments through visual data obtained by a network of cameras. It has also developed approaches to handle problems related to surveillance in order to reduce the impact caused by the propagation of errors along the chain of problems of interest and to increase the speed of the methods...Ler Mais

ARDOP: Robust and Discriminatory Approaches to People Observation

February 14, 2013

The automatic understanding of human activities in videos allows the monitoring of environments based on the analysis of the interaction between individuals and their behaviors. In this way, new technologies for accident prevention and for the identification of suspicious behavior can be developed. This project aimed at handling problems related to people observation focusing on robust and discriminatory approaches...Ler Mais