A Transformer-Based Model for Predicting Human Gaze Behavior in Videos
A novel reinforcement learning algorithm using transformers can accurately predict human gaze behavior in third-person view videos, enabling automation of video understanding tasks that rely on human gaze input.
Reducing Annotation Cost for Video Instance Segmentation with Point Supervision
This work introduces a point-supervised video instance segmentation framework that can achieve competitive performance compared to fully-supervised methods, by leveraging class-agnostic proposal generation and a spatio-temporal point-based matcher to generate high-quality dense pseudo-labels from sparse point annotations.
Leveraging CLIP's Multimodal Capabilities for Robust Video Highlight Detection
By finetuning the pre-trained CLIP model, we achieve state-of-the-art performance on the video highlight detection task, demonstrating the power of leveraging large-scale multimodal knowledge for specialized video understanding.
Action Detection via a Diffusion-based Image Generation Process
The core message of this paper is that action detection can be effectively tackled by formulating it as a three-image generation problem, where the starting point, ending point, and action-class predictions are generated as images via a diffusion-based framework.
Efficient Temporal Boundary Expansion for Weakly Supervised Video Grounding with Multimodal Large Language Model
Effiziente Erweiterung von zeitlichen Grenzen für schwach überwachte Videobegründung mit multimodellen großen Sprachmodellen.
DiffusionVMR: Video Moment Retrieval and Highlight Detection Model
DiffusionVMR proposes a novel denoising generation framework for joint video moment retrieval and highlight detection, leveraging diffusion models for iterative refinement and improved performance.
IVAC-P2L: Leveraging Irregular Repetition Priors for Video Action Counting
モデルの性能を向上させるために、不規則な繰り返し事前条件を活用するIVAC-P2Lモデルが提案されました。
Neuro-Symbolic Video Search: Enhancing Scene Identification with Temporal Logic Reasoning
Decoupling semantic understanding and temporal reasoning is essential for efficient scene identification.
Evaluating Natural Language Video Localization Performance on Multiple Reliable Videos Pool
Proposing the MVMR task to improve video moment retrieval by addressing limitations in existing methods.
DiffusionVMR: Video Moment Retrieval and Highlight Detection Model
Proposing DiffusionVMR for joint video moment retrieval and highlight detection, leveraging denoising generation to refine boundaries iteratively.
© 2024 by Linnk AI