데이터 과학 스터디 썸네일형 리스트형 [Transformer: Attention Is All You Need] 논문리뷰 Paper: Attention Is All You Need The dominant sequence transduction models are based on complex recurrent or convolutional neural networks in an encoder-decoder configuration. The best performing models also connect the encoder and decoder through an attention mechanism. We propose a new arxiv.org Abstract 논문 발표 당시 주도하던 *sequence transduction 모델은 encoder와 decoder를 포함한 복잡한 recurrent (RNN) 혹은 conv.. 더보기 [Generative Adversarial Nets] 논문 리뷰 Paper: Generative Adversarial Networks We propose a new framework for estimating generative models via an adversarial process, in which we simultaneously train two models: a generative model G that captures the data distribution, and a discriminative model D that estimates the probability that arxiv.org Abstract 논문에서는 적대적과정(adversarial process)을 통해 generative model(생성모델)을 추정하는 새로운 방식을 제시한다. 생성 모.. 더보기 [Fully Convolutional Networks for Semantic Segmentation] 논문 리뷰 Fully Convolutional Networks for Semantic Segmentation Paper: Fully Convolutional Networks for Semantic Segmentation Convolutional networks are powerful visual models that yield hierarchies of features. We show that convolutional networks by themselves, trained end-to-end, pixels-to-pixels, exceed the state-of-the-art in semantic segmentation. Our key insight is to build arxiv.org Sementic Segme.. 더보기 mAP(mean Average Precision) *mAP란 mean Average Precision으로 Object Detection 분야에서 모델의 성능을 평가하는데 주로 사용되는 평가지표이다. 이를 이해하기 위해서는 precision(정밀도), recall(재현율), IoU에 대해서 먼저 알아야 한다. 먼저 아래 Confusion Matrix를 살펴보자. True Positive: 예측한 값이 참인데, 실제로도 참인 경우 False Positive: 예측한 값이 참인데, 실제로는 거짓인 경우 False Negative: 예측한 값이 거짓인데, 실제로는 참인 경우 True Positive: 예측한 값이 거짓인데, 실제로도 거짓인 경우 이 때, Precision(정밀도)은 아래 수식과 같이, TP/TP+FP(참이라고 예측한 것 중에 실제로 참인 비율).. 더보기 [Faster R-CNN]논문 리뷰 Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks Reference https://arxiv.org/abs/1506.01497 Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks State-of-the-art object detection networks depend on region proposal algorithms to hypothesize object locations. Advances like SPPnet and Fast R-CNN have reduced the running time of these detection.. 더보기 이전 1 2 3 4 다음
