SOTAVerified

Activity Recognition

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.

Source: Learning Latent Sub-events in Activity Videos Using Temporal Attention Filters

Papers

Recently AddedMost HypedMost ActiveNeeds VerificationMost Verified

Showing 1–25 of 1322 papers

TitleStatusHype
ALS-HAR: Harnessing Wearable Ambient Light Sensors to Enhance IMU-based Human Activity RecogntionCode3
HARDVS: Revisiting Human Activity Recognition with Dynamic Vision SensorsCode3
SenseFi: A Library and Benchmark on Deep-Learning-Empowered WiFi Human SensingCode2
Self-Supervised Contrastive Pre-Training For Time Series via Time-Frequency ConsistencyCode2
Deep learning for time series classificationCode2
Class-incremental Learning for Time Series: Benchmark and EvaluationCode2
NeuFlow: Real-time, High-accuracy Optical Flow Estimation on Robots Using Edge DevicesCode2
Human Activity Recognition using RGB-Event based Sensors: A Multi-modal Heat Conduction Model and A Benchmark DatasetCode2
SensorLLM: Human-Intuitive Alignment of Multivariate Sensor Data with LLMs for Activity RecognitionCode2
WiMANS: A Benchmark Dataset for WiFi-based Multi-user Activity SensingCode2
AutoFi: Towards Automatic WiFi Human Sensing via Geometric Self-Supervised LearningCode2
Convolutional Tensor-Train LSTM for Spatio-temporal LearningCode1
DANA: Dimension-Adaptive Neural Architecture for Multivariate Sensor DataCode1
CubeLearn: End-to-end Learning for Human Motion Recognition from Raw mmWave Radar SignalsCode1
3D Human Shape and Pose from a Single Low-Resolution Image with Self-Supervised LearningCode1
COMPOSER: Compositional Reasoning of Group Activity in Videos with Keypoint-Only ModalityCode1
Contrastive Learning with Cross-Modal Knowledge Mining for Multimodal Human Activity RecognitionCode1
DATTA: Domain-Adversarial Test-Time Adaptation for Cross-Domain WiFi-Based Human Activity RecognitionCode1
Bridge-Prompt: Towards Ordinal Action Understanding in Instructional VideosCode1
Challenges in Multi-centric Generalization: Phase and Step Recognition in Roux-en-Y Gastric Bypass SurgeryCode1
Autoregressive Adaptive Hypergraph Transformer for Skeleton-based Activity RecognitionCode1
CALDA: Improving Multi-Source Time Series Domain Adaptation with Contrastive Adversarial LearningCode1
COMODO: Cross-Modal Video-to-IMU Distillation for Efficient Egocentric Human Activity RecognitionCode1
Comparing Self-Supervised Learning Techniques for Wearable Human Activity RecognitionCode1
BASAR:Black-box Attack on Skeletal Action RecognitionCode1
Show:102550
← PrevPage 1 of 53Next →
All datasetsRWF-2000Stanford40First-Person Hand Action BenchmarkSelf-stimulatory Behavior Dataset

Benchmark Results

#ModelMetricClaimedVerifiedStatus
1Structured Keypoint PoolingAccuracy93.4—Unverified
2Semi-Supervised Hard Attention (SSHA); pretrained on Deepmind Kinetics datasetAccuracy90.4—Unverified
3Human Skeletons + Change DetectionAccuracy90.25—Unverified
4Separable Convolutional LSTMAccuracy89.75—Unverified
5SPIL ConvolutionAccuracy89.3—Unverified
6Flow Gated NetworkAccuracy87.25—Unverified
#ModelMetricClaimedVerifiedStatus
1FocusCLIPTop-3 Accuracy (%)10.47—Unverified
2CLIPTop-3 Accuracy (%)6.49—Unverified
#ModelMetricClaimedVerifiedStatus
1Boutaleb et al.1:1 Accuracy97.91—Unverified
#ModelMetricClaimedVerifiedStatus
1all-landmark-modelActivity Recognition0.76—Unverified