TaYing

Distill Fine-Grained Features Without SupervisionFigure 1. Self-supervised learning is an important step to true artifical intelligence. Image retrieved from Unsplash.Previously, I have written several articles briefly discussing self-supervised learning and, in particular, contrastive learning. What was not yet covered, however, was a concurrent branch of self-supervised approach using interactions of multiple networks that seems to emerge and excel recently. As of today, one of the state-of-the-art training methods is a…

When a neural network uses so much data it becomes “universal”Figure 1. Billions of images are stored on clouds across the internet. Using them for machine learning could potentially be extremely helpful. Image retrieved from https://unsplash.com/photos/M5tzZtFCOfs.Pre-defined classes and categories: this is the limitation where new classes can only be classified by machine learning and neural networks after retraining. For a period of time, this retraining and fine-tuning procedure have almost become “standard” — it is…

Image Denoising with No Data and a Random NetworkFigure 1. DIP Pipeline. A single image is used for training, and the aim is to reconstruct the image from the noise. Eventually the network learns to reconstruct a denoised version of the image. Image created by author.Deep learning and neural networks have been tightly associated with big data. Whether it is image classification of language translation, you almost always require a vast quantity of data to boost the task accuracy for the model to be applicable to real-world…

Can they one day replace photos?Figure 1. NeRF Pipeline. Given a large set of images, NeRF learns to implicitly represent the 3D shape, such that new views can later on be synthesised. Image retrieved from the original NeRF paper by Mildenhall et al.The deep learning era began through the advancements it brought in traditional 2D image-recognition tasks such as classifications, detections, and instance segmentations. As the techniques matured, the research in deep-learning-based computer vision has been shifted towards…

The frontier of deep learning techniques in computer visionWhat is currently going on in the computer vision community? If you are an avid computer vision enthusiast like me, here are several of my favourite papers in 2021–2022 that I believe will have massive impact going into 2022.Disclaimer: These papers are what I think of as “foundational” papers. There are numerous papers extending beyond them with great insights, and I will gradually add the links here for reference.Figure 1. Masked Autoencoder pipeline. Image…