Paper Image

Efficient video object annotation

Published on:

8 November 2023

Primary Category:

Computer Vision and Pattern Recognition

Paper Authors:

Thanos Delatolas,

Vicky Kalogeiton,

Dim P. Papadopoulos


Key Details

EVA-VOS selects which frame and annotation type to provide, rather than annotating all frames

Agent maximizes annotation impact on model while minimizing human time

Experiments show 3.5x faster annotation than standard manual approach

Frame selection method achieves state-of-the-art performance

Significant gains in annotation time versus other methods

AI generated summary

Efficient video object annotation

This paper proposes EVA-VOS, a new framework for efficiently annotating objects in videos using segmentation masks. It introduces an agent that selects which video frame and annotation type a human should provide to maximize annotation impact while minimizing time. Experiments show EVA-VOS produces masks close to human accuracy 3.5x faster than traditional manual annotation, and outperforms other methods in annotation time.

Answers from this paper


No comments yet, be the first to start the conversation...

Sign up to comment on this paper

Sign Up