PK PyTorchKR Papers
Papers with Code · CC BY-SA 4.0

Activity Recognition

1,654 평가 기록 208 datasets

Human Activity Recognition is the problem of identifying events performed by humans given a video input. It is formulated as a binary (or multiclass) classification problem of outputting activity class labels. Activity Recognition is an important problem with many societal applications including smart surveillance, video search/retrieval, intelligent robots, and other monitoring systems.

<span class="description-source">Source: Learning Latent Sub-events in Activity Videos Using Temporal Attention Filters </span>

개요는 각 항목의 일부입니다. 위 탭에서 전체 기록을 페이지별로 탐색할 수 있습니다.

벤치마크

Papers with Code 아카이브의 데이터셋별 기록 예시입니다. 평가 조건과 지표 방향이 검증되지 않아 순위를 의미하지 않습니다.

Dataset기록된 모델 예시MetricPaperCode
Hulk(Finetune, ViT-L)
Accuracy (CS): 94.3
논문 코드
MVD (Kinetics400 pretrain, ViT-H, 16 frame)
Top-1 Accuracy: 77.3Top-5 Accuracy: 95.7Parameters: 633GFLOPs: 1192x6
논문 코드
FTP-UniFormerV2-L/14
3-fold Accuracy: 99.7
논문 -
ProtoGCN
Accuracy (Cross-Subject): 90.9Accuracy (Cross-Setup): 92.2Ensembled Modalities: 6
논문 코드
VideoMAE V2-g
Average accuracy of 3 splits: 88.7
논문 코드
InternVideo
Top 1 Accuracy: 70.0
논문 코드
VideoMAE V2-g
3-fold Accuracy: 99.6
논문 코드
MVD (ViT-B)
Top-1 Accuracy: 79.7Pre-Training Dataset: Kinetics400Frozen: false
논문 코드
Structured Keypoint Pooling (PPNv2 skeletons+objects)
Accuracy: 52.3
논문 -
LART (Hiera-H, K700 PT+FT)
mAP: 45.1
논문 코드
LLaVAction
Action@1: 58.3Verb@1: 76Noun@1: 69
논문 코드
DSCNet (RGB + Pose)
Accuracy (CS): 97.4Accuracy (CV): 99.4
논문 코드

데이터셋

208개
NTU RGB+DSomething-Something V2UCF101NTU RGB+D 120HMDB-51Something-Something V1UCF101HMDB51Kinetics-Skeleton datasetAVA v2.2EPIC-KITCHENS-100NTU RGB+D

논문

라이브러리

Papers with Code 리더보드 row에 연결된 구현 저장소입니다.

github.com/anish-lakkapragada/ActivRecognition-Autism-Diagnosis 1 논문
github.com/atmguille/Violence-Detection-With-Human-Skeletons 1 논문
github.com/mchengny/RWF2000-Video-Database-for-Violence-Detection 1 논문
github.com/Zedd1558/TwoStreamSepConvLSTM_ViolenceDetection 1 논문

하위 태스크

AR Action Recognition 172 데이터셋 · 1,538 rows MA Multimodal Activity Recognition 10 데이터셋 · 37 rows GA Group Activity Recognition 2 데이터셋 · 18 rows HA Human action generation 7 데이터셋 · 18 rows HA Human Activity Recognition 9 데이터셋 · 17 rows EA Egocentric Activity Recognition 2 데이터셋 · 13 rows MI Muscle-Computer Interfaces (MCIs) 1 데이터셋 · 2 rows RA Recognizing And Localizing Human Actions 1 데이터셋 · 1 rows CA Concurrent Activity Recognition 0 데이터셋 · 0 rows CA Cross-Domain Activity Recognition 0 데이터셋 · 0 rows SR Stay Region Extraction 0 데이터셋 · 0 rows