Built independently by an author, for readers. Read the story and support ChapterPal
ChapterPal Logo
Log inSign up

keyword

frame-level representations

Frame-level representations are feature vectors or embeddings that capture the visual and spatial content of individual frames within a video sequence. In video analysis and computer vision architectures, these representations are typically generated by processing each frame through an image-based feature extractor, such as a convolutional neural network or a spatial vision transformer. They capture static scene information, object attributes, and appearance cues at discrete points in time. By serving as an intermediate stage between raw frame inputs and video-level modeling, frame-level representations provide a structured sequence of spatial features that subsequent temporal layers or pooling mechanisms can aggregate to understand motion, temporal dynamics, and overall video semantics.

1 item

ViViT: A Video Vision Transformer

ViViT: A Video Vision Transformer

Anurag Arnab, Mostafa Dehghani, Georg Heigold, Chen Sun, Mario Lučić, Cordelia Schmid

OrganizationsGoogle

Why you should read this

Introduces ViViT, an efficient pure-transformer framework for video classification that factorizes spatio-temporal dimensions and leverages pretrained image models to surpass standard 3D convolutional networks across major benchmarks.

We present pure-transformer based models for video classification, drawing upon the recent success of such models in image classification. Our model extracts spatio-temporal tokens from the input video, which are then encoded by a series of transformer layers. In order to handle the long sequences of tokens encountered in video, we propose several, efficient variants of our model which factorise the spatial- and temporal-dimensions of the input. Although transformer-based models are known to only be effective when large training datasets are available, we show how we can effectively regularise the model during training and leverage pretrained image models to be able to train on comparatively small datasets. We conduct thorough ablation studies, and achieve state-of-the-art results on multiple video classification benchmarks including Kinetics 400 and 600, Epic Kitchens, Something-Something v2 and Moments in Time, outperforming prior methods based on deep 3D convolutional networks. To facilitate further research, we release code at this https URL

Added

2026-09-12