PhD/Postdoctoral researcher position

Weakly supervised deep learning for face and facial attributes detection

(FAPESP fellowship)

Rogerio S. Feris, IBM TJ Watson Research Center, NY, US

Roberto M. Cesar-Jr., University of São Paulo - USP, São Paulo, Brazil

(Published on August / 2014)


    Face and facial attributes, such as beard or glasses, detection play a fundamental role in different computer vision applications. Despite the efforts and research advances in computer vision, the detectors operate in a manner more reliable for frontal poses without occlusion and without the presence of accessories like sunglasses and hat. On the other hand, deep learning techniques have recently shown exceptional results in different problems. The successful deep learning techniques is largely due to two factors: high dimensionality classifiers and large datasets to train them, one aspect of the big data phenomenon. A challenge to use this formula lies in obtaining large sets of annotated training data (with labels) for supervised training. This postdoctoral project aims to develop an original method to generate a large scale database with few annotation automatically and strategies to use these data to train, in a weakly supervised way, a face and facial attributes detector using deep learning techniques.

    We are looking for PhD/Post-doctoral candidates to apply for a FAPESP fellowship (see the conditions below) to develop research based on these ideas. The project will be developed under the Postdoc co-supervision of Rogerio S. Feris and Roberto M. Cesar Jr. The research will be carried out at USP, São Paulo Brazil.

    The successful candidate will actively participate in R&D and also in team activities such as organization of seminars, short courses, study group meetings and co-supervision of undergraduate and graduate students. All these activities are related to the Research Support Centers NAP e-Science of the University of São Paulo.

Required qualifications / skills
  • Undergraduate/Graduate degree with strong background in computation (Computer Science, Engineering, Physics, Math or proven ability to design and implement computer algorithms/systems)
  • Research experience in one or more of the following areas: signal/image processing, image analysis, pattern recognition/analysis, machine learning, multivariate data analysis
  • Oral and written communication skills (English)
    Additional desired qualifications/skills
    • Knowledge of statistical data analysis
    • Proven ability to design data analysis experiments and evaluate their results
    • Programming skills
Application procedure and schedule
    Application deadline: We accept applications on a rolling basis.
    Required documents:
    • CV
    • Summary of the MSc/doctoral thesis and other relevant works
    • Names, positions and emails of two people that may provide you with references

    Fellowship conditions: The selected candidate will apply for a fellowship from FAPESP (São Paulo Research Foundation) with the one of the following conditions: PhD (3-4 years, fellowship US$ 16K/year plus US$ 4.9K/year for  for travel expenses such as attending to conferences); Postdoc (2-4 years, fellowship US$ 30K/year plus US$ 4.5K/year for travel expenses such as attending to conferences). All fellowships may cover expenses for moving to São Paulo/Brasil (including flight tickets). PhD and Postdoc fellowships may apply for an extra period (4 months to 1 year) to do part of the research in any top research university abroad (FAPESP BEPE fellowship). Medical care at the University of São Paulo hospital is also available.

    Please email applications and inquiries to Prof. Roberto M. Cesar Jr.:
      Subject: NAP/USP Postdoc application

IME/USP - Institute of Mathematics and Statistics, USP, São Paulo, Brazil