AVA

The AVA dataset is a benchmark used to evaluate Video Action Detection (VAD) by providing annotated video clips that contain diverse human actions, allowing models to localize and categorize these actions effectively.

Papers using AVA (4)

Language Integration in Fine-Tuning Multimodal Large Language Models for Image-Based Regression2025Group Relative Augmentation For Data Efficient Action Detection2025Mug-STAN: Adapting Image-Language Pretrained Models for General Video Understanding2023JoVALE: Detecting Human Actions in Video Using Audiovisual and Language Contexts2024