Previous research observed accuracy degradation when replacing the atten...
We introduce OpenFlamingo, a family of autoregressive vision-language mo...
Image captioning is conventionally formulated as the task of generating
...
Deep neural networks have reached human-level performance on many comput...
Does progress on ImageNet transfer to real-world datasets? We investigat...
Vision Transformers convert images to sequences by slicing them into pat...
Explainability has become a central requirement for the development,
dep...
Today's computer vision models achieve human or near-human level perform...
We revisit the challenging problem of training Gaussian-Bernoulli restri...
Many contrastive representation learning methods learn a single global
r...
Forward gradient learning computes a noisy directional gradient and is a...
Open-vocabulary models like CLIP achieve high accuracy across many image...
In this work, we study different approaches to self-supervised pretraini...
Semantic segmentation labels are expensive and time consuming to acquire...
Recent progress in Medical Artificial Intelligence (AI) has delivered sy...
The conventional recipe for maximizing model accuracy is to (1) train
mu...
Recent work has uncovered a striking phenomenon in large-capacity neural...
Pre-training (PT) followed by fine-tuning (FT) is an effective method fo...
Understanding the operation of biological and artificial networks remain...
Convolutional neural networks (CNNs) have so far been the de-facto model...
Self-supervised pretraining followed by supervised fine-tuning has seen
...
Self-supervised representation learning has witnessed significant leaps
...
Effective training of deep neural networks can be challenging, and there...
It is common to use the softmax cross-entropy loss to train neural netwo...
A key factor in the success of deep neural networks is the ability to sc...
One paradigm for learning from few labeled examples while making best us...
This paper presents SimCLR: a simple framework for contrastive learning ...
The problem of adversarial examples has highlighted the need for a theor...
After a large "teacher" neural network has been trained on labeled data,...
Convolutional neural networks are among the most successful architecture...
Recent work has indicated that, unlike humans, ImageNet-trained CNNs ten...
Although deep convolutional neural networks achieve state-of-the-art
per...
The generalization and learning speed of a multi-class neural network ca...
To generalize to novel visual scenes with new viewpoints and new object
...
Recent work has sought to understand the behavior of neural networks by
...
Transfer learning is a widely used method to build high performing compu...
Robust risk minimisation has several advantages: it has been studied wit...
Transfer learning has become a cornerstone of computer vision with the a...