👉 Learn how to build AI systems for medical imaging domain by leveraging tools and techniques that I share with you! | 💡 The newsletter is read by people from: Nvidia, Baker Hughes, Harvard, NYU, Columbia University, University of Toronto and more!
Unusual Data Augmentation Techniques!
Hello Reader,
Welcome to another edition of PYCAD newsletter where we cover interesting topics in Machine Learning and Computer Vision applied to Medical Imaging. The goal of this newsletter is to help you stay up-to-date and learn important concepts in this amazing field! I've got some cool insights for you below ↓
In computer vision, there are so many data augmentation techniques that can be used to regularize your model training.
Some of these techniques are well known, such as: translating an image in 2 directions (x and y axis), rotating an image, shear of an image, …etc.
Here are some other data augmentation techniques that you may not know ↓
MixUp
Mixup is a data augmentation technique that generates a weighted combination of random image pairs from the training data.
​
Given two images and their ground truth labels: (x1,y1),(x2,y2), a synthetic training example (x_, y_) is generated as:
​
x_ = lambda * x1 + (1 - alpha) * x2
y_ = lambda * y1 + (1 - alpha) * y2
​
lambda is a value that's sampled from a Beta distribution.
​
Below is an illustration of this data augmentation technique.
More on MixUp can be found here.
CutMix
CutMix is an image data augmentation strategy.
​
Instead of simply removing pixels, we replace the removed regions with a patch from another image.
​
The ground truth labels are also mixed proportionally to the number of pixels of combined images.
​
The added patches further enhance localization ability by requiring the model to identify the object from a partial view.
​
Below you can see a summary of the improvement that this technique has introduced on ImageNet classification.
More on CutMix can be found here.
That's it for this week's edition, I hope you enjoyed it!
Machine Learning for Medical Imaging
by Nour Islam Mokhtari from pycad.co
👉 Learn how to build AI systems for medical imaging domain by leveraging tools and techniques that I share with you! | 💡 The newsletter is read by people from: Nvidia, Baker Hughes, Harvard, NYU, Columbia University, University of Toronto and more!
Hi Reader,, Welcome to the PYCAD newsletter, where every week you receive doses of machine learning and computer vision techniques and tools to help you learn how to build AI solutions to empower the most vulnerable members of our society, patients. TotalSegmentator : Whole Body Segmentation at your Fingertips This free tool available online can do full body segmentation, it's called TotalSegmentator. I have already mentioned this tool in a previous edition of the newsletter, but in this...
Hello Reader, Welcome to another edition of PYCAD newsletter where we cover interesting topics in Machine Learning and Computer Vision applied to Medical Imaging. The goal of this newsletter is to help you stay up-to-date and learn important concepts in this amazing field! I've got some cool insights for you below ↓ A Medical Imaging Expert Told Me This Recently I saw a post on LinkedIn where a medical imaging expert showcased his work of segmenting the lungs and its bronchial trees. You can...
Hello Reader, Welcome to another edition of PYCAD newsletter where we cover interesting topics in Machine Learning and Computer Vision applied to Medical Imaging. The goal of this newsletter is to help you stay up-to-date and learn important concepts in this amazing field! I've got some cool insights for you below ↓ How we helped accelerate inference time for a client's AI product Below is a screenshot of a benchmark we did for a client of ours. The goal was to accelerate inference time. This...