Supervised Contrastive Learning (Prannay Khosla et al.) is a training methodology that outperforms supervised training with cross-entropy on classification tasks.
The idea is that training models using Supervised Contrastive Learning (SCL) can make the model encoder learn better class representation from the samples, this should lead to better generalization and robustness to image and label corruption.
In this article you will learn what is, and how supervised contrastive learn works, you will see the code implementation, an use case application and finally a comparison between SCL and regular cross-entropy.
Clusters of points belonging to the same class are pulled together…
If you want to train your models with Tensorflow in the most efficient way you probably should use TFRecords and the Tensorflow data module to build your pipelines, but depending on the requirements and constraints of your applications, using them might be necessary not and an option, the good news is that Tensorflow has made both of them pretty clean and easy to use.
In this article, we will go through a simple yet efficient way of building pipelines with complex combinations of data augmentation using the Tensorflow data module.
A while ago (early 2020), Kaggle added TPUs as a hardware option to be used on its kernel environment, shortly after, they launched the competition “Flower Classification with TPUs” and challenged the community to use a large dataset of images and extract the most of this powerful hardware.
The objective of this article is to provide basic knowledge about the integration between Tensorflow and TPUs so that you can build a good image classifier baseline that can take advantage of the TPU processing power.
TPU is short for “Tensor processing unit”, TPUs are powerful hardware accelerators specialized in deep learning…
Using deep learning to identify melanomas from skin images and patient meta-data
Kaggle, SIIM, and ISIC hosted the SIIM-ISIC Melanoma Classification competition on May 27, 2020, the goal was to use image data from skin lesions and the patients meta-data to predict if the skin image had a melanoma or not, here is a small introduction to the task from the hosts:
Skin cancer is the most prevalent type of cancer. Melanoma, specifically, is responsible for 75% of skin cancer deaths, despite being the least common skin cancer. The American Cancer Society estimates over 100,000 new melanoma cases will be…
Data Scientist | Google Developer Expert on Machine Learning