Motivation One of the biggest challenges during designing new neural network architectures is time. During real-world workflows, one often trains many different neural networks during the experimentation and design process. This is a wasteful process in which each new model is trained from scratch. In a typical workflow, one trains multiple models, with each model … Continue reading Paper Explanation: Net2Net – Accelerating Learning via Knowledge Transfer

# Category: Deep Learning

## Paper Explanation: Binarized Neural Networks: Training Neural Networks with Weights and Activations Constrained to +1 or −1

Motivation What if you want to do some real-time face detection/recognition using a deep learning system running on a pair of glasses? What if you want your alarm clock to be able to record and analyze your sleep and conditions around you and come up with the most optimal way of waking you up each … Continue reading Paper Explanation: Binarized Neural Networks: Training Neural Networks with Weights and Activations Constrained to +1 or −1

## Paper Explanation – Deep Residual Learning for Image Recognition (ResNet)

This paper is the outcome when Microsoft finally released the beast! The ResNet "slayed" everything, and won not one, not two, but five competitions; ILSVRC 2015 Image Classification, Detection and Localization, and COCO 2015 detection and segmentation. Problems the Paper Addressed The paper analysed what was causing the accuracy of deeper networks to drop as … Continue reading Paper Explanation – Deep Residual Learning for Image Recognition (ResNet)

## Paper Explanation: Going Deeper with Convolutions (GoogLeNet)

This is the only paper I know of that references a meme! Not only this but this model also became the state of the art for classification and detection in the ImageNet Large-Scale Visual Recognition Challenge 2014 (ILSVRC14). Problems the Paper Addressed To create a deep network with high accuracy while keeping computation low. Google … Continue reading Paper Explanation: Going Deeper with Convolutions (GoogLeNet)

## Paper Explanation: Very Deep Covolutional Networks for Large-Scale Image Recognition (VGGNet)

This is probably the most simple looking and straightforward network ever. Simple yet powerful; wining the ImageNet 2014 Localisation competition and coming second in the Classification track! Problems the Paper Addressed To show that making a network deeper improves its accuracy and also multiple small filters are better than a single large filter when both … Continue reading Paper Explanation: Very Deep Covolutional Networks for Large-Scale Image Recognition (VGGNet)

## Paper Explanation : ImageNet Classification with Deep Convolutional Neural Networks (AlexNet)

AlexNet famously won the 2012 ImageNet LSVRC-2012 competition by a large margin (15.3% VS 26.2% (second place) top-5 test error rates). This started the era of deep learning, bringing neural networks back into the spotlight! Problems the paper addressed To show that it is possible to successfully train a deep CNN with a large number … Continue reading Paper Explanation : ImageNet Classification with Deep Convolutional Neural Networks (AlexNet)

## Paper Explanation: Faster R-CNN Towards Real-Time Object Detection with Region Proposal Networks

What Led to Faster R-CNN? Even though networks like Fast R-CNN achieve a really high accuracy, they are very slow to be of practical use in real time. The reason for such slow speeds is the region proposal step in the architecture! Algorithms, like Selective Search (used in Fast R-CNN) take around 2 seconds per … Continue reading Paper Explanation: Faster R-CNN Towards Real-Time Object Detection with Region Proposal Networks