목록CV (3)
On the journey of
[논문읽기] GAN Compression ; Efficient Architectures for Interactive Conditional GANs (2020, CVPR)
Original Paper) https://arxiv.org/pdf/2003.08936.pdf (Direct PDF Download) https://arxiv.org/abs/2003.08936v4 GAN Compression: Efficient Architectures for Interactive Conditional GANs Conditional Generative Adversarial Networks (cGANs) have enabled controllable image synthesis for many vision and graphics applications. However, recent cGANs are 1-2 orders of magnitude more compute-intensive than..
[논문읽기] An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale
Original Paper ) https://arxiv.org/abs/2010.11929 An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale While the Transformer architecture has become the de-facto standard for natural language processing tasks, its applications to computer vision remain limited. In vision, attention is either applied in conjunction with convolutional networks, or used to rep arxiv.org 0. Key..
[논문읽기]3D common corruptions and data augmentation
Abstract neural network 트레이닝을 위한 data augmentation 기법들과 동시에 모델의 robustness 평가하기 위한 image transformation set를 제안한다. 주요 차이점은 Common corruption과 같은 기존 접근 방식과는 달리 scene의 기하학이 적용되어 보다 real world와 같은 변환을 만들어내며 semantic corruption들에 대해 소개한다. 이런 변환이, 효율적(실시간 계산)이고, 확장 가능(기존 데이터셋에 적용)하며, robustness를 향상 시킬 수 있음을 보여줌 1. Introduction Real world에 배치된 컴퓨터 비전 모델들은 그들의 training data에서 자연스럽게 발생된 분포로 옮기게 된다. 이러한 ..