Multi-scale 2D Temporal Map Diffusion Models for Natural Language Video Localization
This addresses the challenge of grounding phrases to video segments for video understanding, representing an incremental improvement with a novel method for a known bottleneck.
The paper tackles the problem of Natural Language Video Localization by proposing a method that generates a global 2D temporal map using a conditional denoising diffusion process to capture temporal dynamics, achieving effective results on Charades and DiDeMo datasets.
Natural Language Video Localization (NLVL), grounding phrases from natural language descriptions to corresponding video segments, is a complex yet critical task in video understanding. Despite ongoing advancements, many existing solutions lack the capability to globally capture temporal dynamics of the video data. In this study, we present a novel approach to NLVL that aims to address this issue. Our method involves the direct generation of a global 2D temporal map via a conditional denoising diffusion process, based on the input video and language query. The main challenges are the inherent sparsity and discontinuity of a 2D temporal map in devising the diffusion decoder. To address these challenges, we introduce a multi-scale technique and develop an innovative diffusion decoder. Our approach effectively encapsulates the interaction between the query and video data across various time scales. Experiments on the Charades and DiDeMo datasets underscore the potency of our design.