Zero Shot Action Recognition On Hmdb51

Top-1 Accuracy

Results

Performance results of various models on this benchmark

Comparison Table

Model Name	Top-1 Accuracy
alignment-uniformity-aware-representation	39
victr-video-conditioned-text-representations	51.0
synthetic-sample-selection-for-generalized	35.9
rethinking-zero-shot-action-recognition	38.7
locate-gat-modeling-multi-scale-local-context	50.7
expanding-language-image-pretrained-models	44.6
evaluation-of-output-embeddings-for-fine	13.3
leveraging-temporal-contextualization-for	56.0
transferring-textual-knowledge-for-visual	58.4
multi-task-zero-shot-action-recognition-with	19.7
multimodal-open-vocabulary-video	60.8
objects2action-classifying-and-localizing	15.6
video-text-modeling-with-zero-shot-transfer	58.7
i-know-the-relationships-zero-shot-action	23.2
msqnet-actor-agnostic-action-recognition-with	-
bidirectional-cross-modal-knowledge	61.4
orthogonal-temporal-interpolation-for-zero	64
alternating-gradient-descent-and-mixture-of	59.1
Model 19	18.5
elaborative-rehearsal-for-zero-shot-action	35.3
multimodal-open-vocabulary-video	64.7
alternative-semantic-representations-for-zero	21.8
claster-clustering-with-reinforcement	43.2
cross-modal-representation-learning-for-zero	41.1
zero-shot-action-recognition-with-error	22.6
towards-universal-representation-for-unseen	24.4
match-expand-and-improve-unsupervised	52.3
rethinking-zero-shot-video-classification-end	32.7
ost-refining-text-knowledge-with-optimal	55.9