Zero-Shot Learning
태스크의 하위 분류와 논문·코드·평가·기법·데이터셋을 탐색하세요.
- 하위 태스크
- 7
- 평가표
- 33
- 보존된 평가 기록
- 84
Zero-shot learning (ZSL) is a model's ability to detect classes never seen during training. The condition is that the classes are not known during supervised learning.
Earlier work in zero-shot learning use attributes in a two-step approach to infer unknown classes. In the computer vision context, more recent advances learn mappings from image feature space to semantic space. Other approaches learn non-linear multimodal embeddings. In the modern NLP context, language models can be evaluated on downstream tasks without fine tuning.
Benchmark datasets for zero-shot learning include aPY, AwA, and CUB, among others.
( Image credit: Prototypical Networks for Few shot Learning in PyTorch )
Further readings:
논문
이 태스크와 하위 분야의 관련 논문입니다. 실제 성능 비교는 평가 결과에서 확인하세요.
On human motion prediction using recurrent neural networks
Motion Prediction Under Multimodality with Conditional Stochastic Networks
Attributes2Classname: A discriminative model for attribute-based unsupervised zero-shot learning
From Zero-shot Learning to Conventional Supervised Classification: Unseen Visual Data Synthesis
Action Tubelet Detector for Spatio-Temporal Action Localization
Investigation of Different Skeleton Features for CNN-based 3D Action Recognition
Cascaded Boundary Regression for Temporal Action Detection
Joint Denoising / Compression of Image Contours via Shape Prior and Context Tree
Action Understanding with Multiple Classes of Actors
Improving Facial Attribute Prediction using Semantic Segmentation
Semantic Autoencoder for Zero-Shot Learning
Molecular De Novo Design through Deep Reinforcement Learning
Skeleton-based Action Recognition with Convolutional Neural Networks
An Analysis of Action Recognition Datasets for Language and Vision Tasks
Body Joint guided 3D Deep Convolutional Descriptors for Action Recognition
Monocular Visual Odometry with a Rolling Shutter Camera
Second-order Temporal Pooling for Action Recognition
Context-based Object Viewpoint Estimation: A 2D Relational Approach
Skeleton Boxes: Solving skeleton based action detection with a single deep convolutional neural network
Skeleton based action recognition using translation-scale invariant image mapping and multi-scale deep cnn
Temporal Action Localization by Structured Maximal Sums
Interpretable 3D Human Action Analysis with Temporal Convolutional Networks
Learning to Estimate Pose by Watching Videos
Hide-and-Seek: Forcing a Network to be Meticulous for Weakly-supervised Object and Action Localization
Automatic Discovery, Association Estimation and Learning of Semantic Attributes for a Thousand Categories
UC Merced Submission to the ActivityNet Challenge 2016
Semantically Consistent Regularization for Zero-Shot Recognition
Modeling Temporal Dynamics and Spatial Configurations of Actions Using Two-Stream Recurrent Neural Networks
First-Person Hand Action Benchmark with RGB-D Videos and 3D Hand Pose Annotations
Generalized Rank Pooling for Activity Recognition
Action Representation Using Classifier Decision Boundaries
Two Stream LSTM: A Deep Fusion Framework for Human Action Recognition
Chained Multi-stream Networks Exploiting Pose, Motion, and Appearance for Action Classification and Detection
Unsupervised Action Proposal Ranking through Proposal Recombination
Hidden Two-Stream Convolutional Networks for Action Recognition
출처: Papers with Code · CC BY-SA 4.0. 아카이브 자료와 새로 수집한 논문은 별도로 관리합니다.