no code implementations • 17 Jun 2021 • Levy Chaves, Alceu Bissoto, Eduardo Valle, Sandra Avila
Self-supervised pre-training appears as an advantageous alternative to supervised pre-trained for transfer learning.
2 code implementations • 3 May 2021 • Juliano S. Assine, J. C. S. Santos Filho, Eduardo Valle
In the past few years, mobile deep-learning deployment progressed by leaps and bounds, but solutions still struggle to accommodate its severe and fluctuating operational restrictions, which include bandwidth, latency, computation, and energy.
1 code implementation • 20 Apr 2021 • Alceu Bissoto, Eduardo Valle, Sandra Avila
Despite the growing availability of high-quality public datasets, the lack of training samples is still one of the main challenges of deep-learning for skin lesion analysis.
1 code implementation • 24 Jun 2020 • Arthur Douillard, Eduardo Valle, Charles Ollion, Thomas Robert, Matthieu Cord
Continual learning aims to learn tasks sequentially, with (often severe) constraints on the storage of old learning samples, without suffering from catastrophic forgetting.
1 code implementation • 28 Apr 2020 • Vinicius Ribeiro, Sandra Avila, Eduardo Valle
Segmenting skin lesions images is relevant both for itself and for assisting in lesion classification, but suffers from the challenge in obtaining annotated data.
1 code implementation • ECCV 2020 • Arthur Douillard, Matthieu Cord, Charles Ollion, Thomas Robert, Eduardo Valle
Lifelong learning has attracted much attention, but existing works still struggle to fight catastrophic forgetting and accumulate knowledge over long stretches of incremental learning.
1 code implementation • 23 Apr 2020 • Alceu Bissoto, Eduardo Valle, Sandra Avila
Data-driven models are now deployed in a plethora of real-world applications - including automated diagnosis - but models learned from data risk learning biases from that same data.
no code implementations • 23 Nov 2019 • Juliano S. Assine, Alan Godoy, Eduardo Valle
Despite recent advances in architectures for mobile devices, deep learning computational requirements remains prohibitive for most embedded devices.
no code implementations • 29 Oct 2019 • Alceu Bissoto, Eduardo Valle, Sandra Avila
Generative Adversarial Networks fostered a newfound interest in generative models, resulting in a swelling wave of new works that new-coming researchers may find formidable to surf.
1 code implementation • 6 Jun 2019 • Vinicius Ribeiro, Sandra Avila, Eduardo Valle
We also evaluate how conditioning the ground truths using different (but very simple) algorithms may help to enhance agreement and may be appropriate for some use cases.
1 code implementation • 29 Apr 2019 • Fábio Perez, Sandra Avila, Eduardo Valle
We evaluate that claim for melanoma classification, over 9 CNNs architectures, in 5 sets of splits created on the ISIC Challenge 2017 dataset, and 3 repeated measures, resulting in 135 models.
1 code implementation • 18 Apr 2019 • Alceu Bissoto, Michel Fornaciali, Eduardo Valle, Sandra Avila
We fed models with additional clinically meaningful information, which failed to improve the results even slightly, suggesting the destruction of cogent correlations.
2 code implementations • 8 Feb 2019 • Alceu Bissoto, Fábio Perez, Eduardo Valle, Sandra Avila
Skin cancer is by far the most common type of cancer.
1 code implementation • 5 Sep 2018 • Fábio Perez, Cristina Vasconcelos, Sandra Avila, Eduardo Valle
In this work, we investigate the impact of 13 data augmentation scenarios for melanoma classification trained on three CNNs (Inception-v4, ResNet, and DenseNet).
no code implementations • 25 Aug 2018 • Alceu Bissoto, Fábio Perez, Vinícius Ribeiro, Michel Fornaciali, Sandra Avila, Eduardo Valle
This extended abstract describes the participation of RECOD Titans in parts 1 to 3 of the ISIC Challenge 2018 "Skin Lesion Analysis Towards Melanoma Detection" (MICCAI 2018).
1 code implementation • 12 Jun 2018 • George Gondim-Ribeiro, Pedro Tabacof, Eduardo Valle
Adversarial attacks are malicious inputs that derail machine-learning models.
1 code implementation • 1 Nov 2017 • Eduardo Valle, Michel Fornaciali, Afonso Menegola, Julia Tavares, Flávia Vasques Bittencourt, Lin Tzy Li, Sandra Avila
We use the exhaustive trials to simulate sequential decisions and ensembles, with and without the use of privileged information from the test set.
2 code implementations • 22 Mar 2017 • Afonso Menegola, Michel Fornaciali, Ramon Pires, Flávia Vasques Bittencourt, Sandra Avila, Eduardo Valle
Knowledge transfer impacts the performance of deep learning -- the state of the art for image classification tasks, including automated melanoma screening.
4 code implementations • 14 Mar 2017 • Afonso Menegola, Julia Tavares, Michel Fornaciali, Lin Tzy Li, Sandra Avila, Eduardo Valle
This extended abstract describes the participation of RECOD Titans in parts 1 and 3 of the ISIC Challenge 2017 "Skin Lesion Analysis Towards Melanoma Detection" (ISBI 2017).
1 code implementation • 5 Dec 2016 • Ramon Oliveira, Pedro Tabacof, Eduardo Valle
We compare the following candidate neural network models: Maximum Likelihood, Bayesian Dropout, OSBA, and --- for MNIST --- the standard variational approximation.
1 code implementation • 1 Dec 2016 • Pedro Tabacof, Julia Tavares, Eduardo Valle
We find that autoencoders are much more robust to the attack than classifiers: while some examples have tolerably small input distortion, and reasonable similarity to the target image, there is a quasi-linear trade-off between those aims.
no code implementations • 5 Sep 2016 • Afonso Menegola, Michel Fornaciali, Ramon Pires, Sandra Avila, Eduardo Valle
Deep learning is the current bet for image classification.
1 code implementation • 11 May 2016 • Micael Carvalho, Matthieu Cord, Sandra Avila, Nicolas Thome, Eduardo Valle
In recent years, deep architectures have been used for transfer learning with state-of-the-art performance in many datasets.
2 code implementations • 14 Apr 2016 • Michel Fornaciali, Micael Carvalho, Flávia Vasques Bittencourt, Sandra Avila, Eduardo Valle
In this paper we survey, analyze and criticize current art on automated melanoma screening, reimplementing a baseline technique, and proposing two novel ones.
no code implementations • 20 Nov 2015 • Otávio A. B. Penatti, Sandra Avila, Eduardo Valle, Ricardo da S. Torres
Results for image classification show that as visual dictionaries are based on low-level visual appearances, visual diversity is more important than semantic diversity.
2 code implementations • 19 Oct 2015 • Pedro Tabacof, Eduardo Valle
Adversarial examples have raised questions regarding the robustness and security of deep neural networks.
no code implementations • 9 Oct 2015 • Thierry Pinheiro Moreira, Mauricio Lisboa Perez, Rafael de Oliveira Werneck, Eduardo Valle
Human facial recognizers perform poorly for dogs (up to 60. 5% accuracy), showing that dog facial recognition is not a trivial extension of human facial recognition.