The present project aims to investigate different strategies to handle the insufficient availability of labeled training data, towards fully exploiting the potential of deep architectures for remote sensing image analysis.
Regarding the application domain, the project aims to employ the solutions to be developed in relevant, concrete problems:
The present project focus on the development of a framework aiming the long-term face tracking under unconstrained scenarios in video sequences. The main idea is to step up the tracking processing by combining a tracking ensemble with the information delivered by a face detector, and also by including a feedback process to provide additional input to the trackers.
Read more at the Project’s site.
The project aims to evaluatie and compare state-of-the-ate Deep Learning techinques for crop mapping in tropical and a temperate regions. It is done using multi-temporal SAR image sequences.
Read more at the Project’s site.
The project aims to research about synthesizing of multispectral optical imagery using Generative Adversarial networks. You can find more information at the Project Site
This project aims to recognize crop types for crop yield estimation using Conditional Random Fields. You can find more information at the Project Site
This project aims to evaluate deep learning techniques applied to Deforestation Detection in the Brazilian Amazon. You can find more information at the Project Site
This project aims to evaluate deep learning techniques applied to domain adaptation for the semantic segmentation of wide areas and time-asynchronous scenes in remote sensing applications. Read more at the Project Site
This research aims to map Savannahs physiognomies over hyperspectral images, using unsupervised deep learning techniques.
This research aims to evaluate and develop fraud detection methods (anti-spoofing) in biometric systems.
The scope mainly involves the face recognition cases.
The project aims to evaluate and develop fraud detection methods (spoofing) in multimodal biometric systems. The scope mainly involves the combination of facial and voice recognition.
Read more at LIVENESS PROJECT
The main objective of the Interimage Cloud Platform is the scientific and technological development in the remote sensing (SR) image analysis, focused on the object based image analysis (GEOBIA) and cloud computing . In terms of scientific development, the main topics of the Project are: (1) the representation and processing of knowledge explicitly represented on an interpretation of SR images; (2) methods of segmentation of SR images, feature extraction and object classification in distributed environments; (3) methods of multitemporal analysis of SR images in distributed environments; (4) methods of advanced visualization of images and objects stored in a distributed way in computer clusters. In terms of technological development, the main objective is the creation of the Interimage Cloud Platform, an innovative platform for GEOBIA, structured on an architecture based on the cloud computing paradigm, in which will be applied the results of the research developed during this project.
The objectives of this research work are:
- Evaluate state-of-the-art technologies for speaker recognition.
- Investigate multimodal fusion techniques.
- Develop and test a prototype of a speaker recognition system.
- Integrate speaker recognition into the face recognition prototype existing in Computer Vision Laboratory – LVC.
The main objective of INTERSAR is to develop in the scope of InterIMAGE, a free open source software platform for GEOBIA applications, methodologies, algorithms and computational tools for automatic Land Use and Land Cover classification upon optical and SAR (synthetic aperture radar) data.
Read more at INTERSAR Project
The general objective of the RADAR MONITOR project is to provide solutions for the systematic monitoring of changes on the Earth’s surface from remote sensing images, with emphasis on data provide by space born SAR (Synthetic-aperture radar) sensors. More precisely, the Project aims to extend the InterIMAGE platform, by incorporating specific software operators for the multitemporal analysis of SAR as well as optical data.
The Brazilian Visualization Network (RVA), whose general coordination is under the Brazilian National Laboratory of Scientific Computation (LNCC), is part of the Brazilian Technology System (SIBRATEC) of the Ministry of Science, Technology and Information (MCTI). The RVA aims at the integration of the academy to foster the innovation and creation of new markets for technological products and services. Beyond LNCC, the RVA involves the following institutions: PUC-Rio, PUC-RS, UFRS, UFPE, UFPB, INPE, USP, UFRJ e UNICAMP, whereby LNCC, UFRJ, PUC-RS, UFPE, USP and PUC-Rio constitute the so called Coordination Nucleus of RVA.
The project “Tools for Open Multi-Risk Assessment using Earth Observation Data” (TOLOMEO) is funded under the Marie Curie International Research Staff Exchange Scheme (PIRSES-GA-2009) with the ultimate goal to establish an international cooperation between partners in Europe and South-America focused on the development of free tools for remotely sensed data analysis.
Read more at TOLOMEO
This project aims to develop technology of video surveillance with automatic facial identification, developing facial recognition techniques from video sequences that integrate methods and tools that allow:
- To detect and track faces in real time based on video sequences,
- To extract frontal facial images in real time based on video sequences, with good quality for recognition,
- To build a prototype of automatic video monitoring.
Read more at Video Surveillance
An open source knowledge based framework for automatic image interpretation.
Read more at InterIMAGE
The project objective is the protection of forest areas by developing a system of satellite monitoring of the hillsides of Rio de Janeiro.
Read more at PIMAR
The face recognition prototype.
Read more at FacePUC