Projects
Machine learning: Ensemble Models
The primary motivation behind this projekt was to get hands on experience with MLFlow,
to get a better understanding of MLOps and hands on experience with the MLFlow API
in general.
My motivation was also to get hands-on experience with docker and containerization, as
this is a sought after skill by employers.
So the ML-model architecture is not going to be the most complex at first, because i am focusing
on simple applications of MLFlow and docker. But over time i will develop the architecture
to become more complex.
- MLflow
- Tensorflow/Pytorch
- Docker & Kubernetes
- Github Actions
Computer vision project
The purpose of the project was to apply computer vision techniques for the
automatic classification of medical images of lymph nodes, aiming to distinguish
between benign- and malignant cells. The goal was to develop a decision support tool for the diagnosis of lymph node cancer.
For this projekt i used the patchcamelyon dataset from Kaggle which consists of 327,680 color images (96x96pxs)
extracted from histopathologic scans of
lymph node sections.
I experimented with developing an autoencoder and a
variational autoencoder to compress- and upscale images and a CNN to classify the output.
After i was satisfied with my baseline
models i started experimenting with different optimizers, loss funktions,
activation funktions and regularization to improve performance.
I have also experimented a bit with transfer learning on the CNN models
that i applied, but this was without using the autoencoder on
the images first for the input.
I got experience using:
- Auto Encoders
- CNN's
- Transfer learning
- Tensorflow
- Keras
- pandas
- numpy
- matplotlib
- sklearn
- Github actions And much more...
