Maxence Boels's Projects
A simple implementation of 3D-Unet on a 3D Prostate Segmentation Task
Keras 3D U-Net Convolution Neural Network (CNN) designed for medical image segmentation
Classification and Segmentation: Diagnose diseases from x-rays and 3D MRI brain images. Predict patient survival rates more accurately using tree-based models. Estimate treatment effects on patients using data from randomized trials.
AMIGO website
Use of Attention Gates in a Convolutional Neural Network / Medical Image Classification and Segmentation
This project attempts to train a Continuous Density Hidden Markov Model (CD-HMM) for speech recognition, and is developed with Matlab software. This objective is reached using the Expectation-Maximization approach using the Baum-Welch equations. The training process uses two steps which are computing the Expectations (E-step) and Maximizing those expectations by re-estimation of the parameters (M-step). The methodology and results are discussed to provide a clear understanding of the motivations and limits of this project.
Shallow neural networks for breast cancer classification as either benign or cancerous tumors.
Public facing notes page
Deep Learning papers reading roadmap for anyone who are eager to learn this amazing tech!
PyTorch code for Vision Transformers training with the Self-Supervised learning method DINO
Temporally smooth online action detection using cycle-consistent future anticipation
having fun with gpt
This project is part of the Computer Vision and Pattern Recognition module at University of Surrey. Students are required to use Matlab to implement this project. The dataset gathers 591 images of 20 classes of different objects e.g. sheep, cow, boat, car and are in RGB color format with different number of pixels. The objective is to develop a visual search engine to retrieve similar images to the one provided by the user. This query image is selected from the data set and thus corresponds to one of the 20 classes.
Predicting lung cancer survival time
About me
Personal website @ https://maxboels.com
A framework for Medical Image Segmentation with Convolutional Neural Networks and Deep Learning
AI Toolkit for Healthcare Imaging
NYU Deep Learning Spring 2021
Two popular classifiers are investigated; the Bayes decision Rule for normally distributed classes and the k-Nearest Neighbour decision rule.
A collection of tutorials and examples for solving and understanding machine learning and pattern classification tasks
Regarding the data, all 628 screening mammograms in this project have been classified as ab-normal by 2 radiologists and thus require a biopsy. Radiologists are cautious during screenings, the consequences of having a false negative push them to send women for biopsy if there is the slight-est doubt for them to have cancer. As explained, this results in a high number of false positive since the βcostβ is lower than having a false negative. The dataset used in this project comes from the OPTIMAM Medical Image Database, which collects NHS Breast Screening Programme (NHSB-SO) images in the UK. A deep learning approach is used to classify abnormal screenings as either malignant or benign cancer with a certain probability. Transfer Learning makes it possible to obtain high performances on small datasets. This project achieved a ROC of 80%, 86% sensitivity, and 77% NPV, which were reached with a pre-trained ResNet50v2, a state-of-the-art neural network optimized through fine-tuning hy-perparameters and data pre-processing.
PyTorch image models, scripts, pretrained weights -- ResNet, ResNeXT, EfficientNet, EfficientNetV2, NFNet, Vision Transformer, MixNet, MobileNet-V3/V2, RegNet, DPN, CSPNet, and more
Retina blood vessel segmentation with a convolutional neural network
This project attempts to solve the problem of speech synthesis of male and female vowels, and is developed with the help of Matlab software. This objective is reached using the Linear Predictive Code approach to estimate the coefficients and formant frequency. Then, vowels are generated by passing an excitation signal through the modeled filter.
Collated a list of useful open-access work related to surgical phase recognition and surgical skills and workflow analysis.