projects

3D Segmentation, Slice Propagation, Uncertainty Quantification

Deep Ensemble
Batch Ensemble
Monte Carlo Dropout
Concrete Dropout
Stochastic Weighted Averaging Gaussian (SWAG)

Supervised methods for 3D anatomy segmentation demonstrate superior performance but are often limited by the availability of annotated data. This limitation has led to a growing interest in self-supervised approaches in tandem with the abundance of available un-annotated data. Slice propagation has emerged as an self-supervised approach that leverages slice registration as a self-supervised task to achieve full anatomy segmentation with minimal supervision. This approach significantly reduces the need for domain expertise, time, and the cost associated with building fully annotated datasets required for training segmentation networks. However, this shift toward reduced supervision via deterministic networks raises concerns about the trustworthiness and reliability of predictions, especially when compared with more accurate supervised approaches. To address this concern, we propose the integration of calibrated uncertainty quantification (UQ) into slice propagation methods, providing insights into the model's predictive reliability and confidence levels. Incorporating uncertainty measures enhances user confidence in self-supervised approaches, thereby improving their practical applicability. We conducted experiments on three datasets for 3D abdominal segmentation using five UQ methods. The results illustrate that incorporating UQ improves not only model trustworthiness, but also segmentation accuracy. Furthermore, our analysis reveals various failure modes of slice propagation methods that might not be immediately apparent to end-users. This study opens up new research avenues to improve the accuracy and trustworthiness of slice propagation methods.

Machine Learning, Binary Classification

Logistic Regression
Random Forest
Decision Tree
Naive Bayes
K Nearest Neighbors
Support Vector Machine
Bagging Decision Trees
AdaBoost
XGBoost
Gradient Boost
Voting Classification
Neural Networks

Image Reconstruction, Computer Vision

Contextual GANs (Generative Adversarial Networks)
GFP-GANs
DeOldify

Criminal investigations often have sketches as the only reference to identify a criminal suspect. Sketch images contain basic face profile information but lack detailing, as is contained in photogenic images. Thus, recognizing actual human faces from those sketches becomes difficult, and it would help to generate a face image from these sketches. The implementation of computer vision into this will reduce the human effort to relate a black and white drawing to actual faces by means of image translation. Previous works have handled the task of transforming viewed sketches into mugshot images, and focused on forensic sketches. In this study, we aim to transform the sketches into realistic facial photographs. With the input sketch the output of common image-to-image translation follows the input edges due to the hard condition imposed by the translation process. Instead, we propose to use sketch as weak constraint, where the output edges do not necessarily follow the input edges. We address this problem using a novel joint image completion approach, where the sketch provides the image context for completing, or generating the output image. We have created a model using contextual Generative Adversarial Networks (GANs). It takes an artistic sketch of a person’s face as the input image, enhances certain features of it and transforms it into a realistic photograph of the face of that person. We have used the CUHK Face Sketch Dataset to train our model.