Joonsu OhReview: PR-302-NeRF: Representing Scenes as Neural Radiance Fields for View SynthesisNeRF created a novel way to construct a 3D model.1 min read·Apr 6, 2022----
Joonsu OhReview: PR-297-DeiTViTs are not as efficient as EfficientNet in terms of its number of parameters.1 min read·Mar 30, 2022----
Joonsu OhReview: PR-284-End-to-End Object Detection with Transformers (DETR)This is the first try that uses a transformer for object detection.1 min read·Mar 29, 2022----
Joonsu OhReview: PR-281-An Image is Worth 16x16 Words (ViT)ViT directly applied transformer architecture on image domain. It divides an image into multiple patches, add positional embedding, then…1 min read·Mar 23, 2022----
Joonsu OhReview: PR-260-Momentum Contrast for Unsupervised Visual Representation LearningThere has been a drastic change in a field of SSL-contrastive learning (SimCLR, MoCo, BYOL…).1 min read·Mar 21, 2022----
Joonsu OhReview: BERTLong time no see! Had a bit of a break and came back finally.3 min read·Dec 4, 2021----
Joonsu OhReview: PR-115-Unsupervised Anomaly Detection with Generative Adversarial NetworksTrain a DCGAN to properly approximate the normal (not Gaussian!) distribution of images of interest. With this, we have one generator that…1 min read·Aug 22, 2021----
Joonsu OhReview: PR-116-Glow: Generative Flow with Invertible 1x1 ConvolutionsWhat is flow based generative models?4 min read·Aug 22, 2021----
Joonsu OhReview: PR-087-Spectral Normalization for Generative Adversarial NetworksOne of the challenges in the study of generative adversarial networks is the instability of its training2 min read·Aug 13, 2021----