1 code implementation • 12 Dec 2023 • Bodong Zhang, Hamid Manoochehri, Man Minh Ho, Fahimeh Fooladgar, Yosep Chong, Beatrice S. Knudsen, Deepika Sirohi, Tolga Tasdizen
On the other hand, acquiring extensive datasets with localized labels for training is not feasible.
Contrastive Learning Histopathological Image Classification +2
1 code implementation • 19 Apr 2023 • Mingzhen Shao, Tolga Tasdizen, Sarang Joshi
This study explores the resilience of a variety of deep homography estimation models to domain shifts, revealing that the network architecture itself is not a contributing factor to this remarkable adaptability.
1 code implementation • 20 Jul 2022 • Ricardo Bigolin Lanfredi, Joyce D. Schroeder, Tolga Tasdizen
We extract snippets from the ET data by associating them with the dictation of keywords and use them to supervise the localization of specific abnormalities.
1 code implementation • 24 Jun 2022 • Bodong Zhang, Beatrice Knudsen, Deepika Sirohi, Alessandro Ferrero, Tolga Tasdizen
Two separate CNNs are used to embed the two views into a joint feature space.
no code implementations • 12 Jan 2022 • Mitra Alirezaei, Tolga Tasdizen
We propose a classification model that does not obfuscate gradients and is robust by construction without assuming prior knowledge about the attack.
1 code implementation • 22 Dec 2021 • Ricardo Bigolin Lanfredi, Ambuj Arora, Trafton Drew, Joyce D. Schroeder, Tolga Tasdizen
The interpretability of medical image analysis models is considered a key research field.
1 code implementation • 29 Sep 2021 • Ricardo Bigolin Lanfredi, Mingyuan Zhang, William F. Auffermann, Jessica Chan, Phuong-Anh T. Duong, Vivek Srikumar, Trafton Drew, Joyce D. Schroeder, Tolga Tasdizen
Furthermore, a small subset of the data contains readings from all radiologists, allowing for the calculation of inter-rater scores.
1 code implementation • 10 Sep 2020 • Ricardo Bigolin Lanfredi, Joyce D. Schroeder, Tolga Tasdizen
To evaluate the alignment with this direction after adversarial training, we apply a metric that uses generative adversarial networks to produce the smallest residual needed to change the class present in the image.
1 code implementation • 4 Jul 2020 • Ricardo Bigolin Lanfredi, Joyce D. Schroeder, Clement Vachet, Tolga Tasdizen
The high complexity of deep learning models is associated with the difficulty of explaining what evidence they recognize as correlating with specific disease labels.
no code implementations • 31 Jan 2020 • Zhaotao Wu, Jia Wei, Wenguang Yuan, Jiabing Wang, Tolga Tasdizen
We introduce the idea of inter-slice image augmentation whereby the numbers of the medical images and the corresponding segmentation labels are increased between two consecutive images in order to boost medical image segmentation accuracy.
1 code implementation • 27 Aug 2019 • Ricardo Bigolin Lanfredi, Joyce D. Schroeder, Clement Vachet, Tolga Tasdizen
We use a conditional generative adversarial network where the generator attempts to learn to shift the output of a regressor through creating disease effect maps that are added to the original images.
1 code implementation • 8 Jul 2019 • Wenguang Yuan, Jia Wei, Jiabing Wang, Qianli Ma, Tolga Tasdizen
Currently, most deep models for multimodal segmentation rely on paired registered images.
no code implementations • 8 Jan 2019 • Ertunc Erdil, Ali Ozgur Argunsah, Tolga Tasdizen, Devrim Unay, Mujdat Cetin
Data driven segmentation is an important initial step of shape prior-based segmentation methods since it is assumed that the data term brings a curve to a plausible level so that shape and data terms can then work together to produce better segmentations.
no code implementations • 3 Sep 2018 • Ertunc Erdil, Sinan Yildirim, Tolga Tasdizen, Mujdat Cetin
In this paper, we propose an efficient pseudo-marginal Markov chain Monte Carlo (MCMC) sampling approach to draw samples from posterior shape distributions for image segmentation.
no code implementations • 3 Jul 2017 • Tolga Tasdizen, Mehdi Sajjadi, Mehran Javanmardi, Nisha Ramesh
We present a neighborhood similarity layer (NSL) which induces appearance invariance in a network when used in conjunction with convolutional layers.
1 code implementation • 14 Aug 2016 • Ting Liu, Miaomiao Zhang, Mehran Javanmardi, Nisha Ramesh, Tolga Tasdizen
We then propose a Bayesian model that combines the supervised and the unsupervised information for probabilistic learning.
Electron Microscopy Image Segmentation Image Segmentation +2
no code implementations • 19 Jul 2016 • Muhammad Usman Ghani, Ertunc Erdil, Sumeyra Demir Kanik, Ali Ozgur Argunsah, Anna Felicity Hobbiss, Inbal Israely, Devrim Unay, Tolga Tasdizen, Mujdat Cetin
We perform cluster analysis on two-photon microscopic images of spines using morphological, shape, and appearance based features and gain insights into the spine shape analysis problem.
no code implementations • 24 Jun 2016 • Fitsum Mesadi, Mujdat Cetin, Tolga Tasdizen
Level set methods are widely used for image segmentation because of their capability to handle topological changes.
no code implementations • 23 Jun 2016 • Fitsum Mesadi, Tolga Tasdizen
The major contributions of this paper include a robust convex decomposition which also results in an efficient part-based shape representation, and a novel shape convexity measure.
no code implementations • NeurIPS 2016 • Mehdi Sajjadi, Mehran Javanmardi, Tolga Tasdizen
Effective convolutional neural networks are trained on large sets of labeled data.
Ranked #14 on Semi-Supervised Image Classification on SVHN, 250 Labels
no code implementations • 9 Jun 2016 • Mehdi Sajjadi, Mehran Javanmardi, Tolga Tasdizen
In this paper we consider the problem of semi-supervised learning with deep Convolutional Neural Networks (ConvNets).
2 code implementations • 4 May 2016 • Mehran Javanmardi, Mehdi Sajjadi, Ting Liu, Tolga Tasdizen
This can be seen as a regularization term that promotes piecewise smoothness of the label probability vector image produced by the ConvNet during learning.
1 code implementation • 24 May 2015 • Ting Liu, Mojtaba Seyedhosseini, Tolga Tasdizen
Starting with over-segmenting superpixels, we use a tree structure to represent the hierarchy of region merging, by which we reduce the problem of segmenting image regions to finding a set of label assignment to tree nodes.
1 code implementation • 30 Dec 2014 • Mehdi Sajjadi, Mojtaba Seyedhosseini, Tolga Tasdizen
Backpropagation, which is used to train artificial neural networks, suffers from the herd effect problem which leads to long training times and limit classification accuracy.
no code implementations • 4 Feb 2014 • Mojtaba Seyedhosseini, Tolga Tasdizen
At each level of the hierarchy, a classifier is trained based on downsampled input images and outputs of previous levels.