Marcela Carvalho

I am actually Research Engineer at Upciti. Our team is working on Video Panoptic Segmention for urban environments.

Previously, I was a PhD student at ONERA working on Deep Learning for Optics and Computer vision. I've received my engineering diploma in Mechatronics in 2014 from the Universidade de Brasilia, followed by a Master diploma in Robotics in 2015 from the Université de Montpellier. I'm passionate by robotics and computer vision. During my undergraduate studies, I've proudly created with some of my colleagues the Droid Team, a robotics team that attends to Brazilian national and International intelligent robotics competitions.

News

2020

[November] Our paper Pix2Point : prédiction monoculaire de scènes 3D par réseaux de neurones hybrides et transport optimal was published to RFIAP 2020. [Paper FR]

2019

[November] Hey! I will defend my Ph.D. thesis on the 25th of November at 2PM, in Paris. If you are interested, join us! Please fill this form so we will have enough food to cry, or to party.

[November] Our paper on Multitask Learning of Height and Semantics From Aerial Images was accepted for IEEE Geoscience and Remote Sensing Letters. [Paper EN] [Code]

[November] We published Technical Report: Co-learning of geometry and semantics for online 3D mapping for our method proposed to ECCV 2018 Workshop: 3D Reconstruction meets Semantics. Our solution performed best on 2 categories of the challenge. [Paper EN]

[October] I've started working at Upciti as Head of AI to develop new solutions for panoptic segmentation in urban environments.

[September-December] Teaching assistant in Programming Introduction at École des Ponts ParisTech with Pascal Monasse PhD..

[April-June] Teaching assistant in Audio Processing Modal at École Polytechnique with Alain Louis-Joseph PhD..

2018

[October] Looking for a 4 to 6 months internship? We have an open offer on Deep neural networks for 3D point cloud prediction from a single image! Check it out!.

[September] We presented our paper on Deep Depth from Defocus: how can defocus blur improve 3D estimation using dense neural networks? at the 3DRW Workshop of ECCV 2018 in Munich. [Paper EN]

[May] Yey! We received the Best Paper Award for our paper on Estimation de profondeur mono-image par réseaux de neurones et flou de défocalisation at RFIAP/CFTP 2018. [Paper FR]

[May] Our paper on Regression losses for deep depth estimation was accepted for ICIP 2018. [Paper EN]

[April-June] Teaching assistant in Audio Processing Modal at École Polytechnique with Alain Louis-Joseph PhD. and Guillaume Pakula.

[April] Oral presentation at JIONC about What Insight Does Defocus Blur Give to Neural Networks for Depth Prediction? [Slides FR] [Video]

2017

[October] Poster presentation at Journée Co-Conception about Depth Estimation with Adversarial Networks. [Résumé FR]

[September] Poster presentation at Gretsi about Depth Estimation with Adversarial Networks. [Poster FR]

[May] IEEE Class Seminar at Universidade de Brasilia about Generative Adversarial Networks and an Application to Robotics.

[March] Poster presentation at JIONC about Deep Monocular Depth Estimation with Non-Conventional Optics. [Poster FR]

2016

[October] Started my PhD on 3D Camera for Depth from Defocus and Machine Learning at ONERA - The French Aerospace Lab with Pauline Trouvé-Peloux PhD., Bertrand Le Saux PhD., Frédéric Champagnat PhD., Andrés Almansa PhD..

2014

[August] Received my diploma in Automation and Control Engineering from the Universidade de Brasilia. My final project was about Teleoperation and control of a NAO humanoid based on a dual-quaternions framework developed under the supervision of Antonio Padilha Lanari Bó PhD. at LARA/UnB. [Manuscript PT]