Sharingan: Extract User Action Sequence from Desktop Recordings

TL;DR

This work represents the first application of VLM-based methods for extracting user action sequences from desktop recordings, contributing new methods, benchmarks, and insights for future research.

Abstract

Video recordings of user activities, particularly desktop recordings, offer a rich source of data for understanding user behaviors and automating processes. However, despite advancements in Vision-Language Models (VLMs) and their increasing use in video analysis, extracting user actions from desktop recordings remains an underexplored area. This paper addresses this gap by proposing two novel VLM-based methods for user action extraction: the Direct Frame-Based Approach (DF), which inputs sampled frames directly into VLMs, and the Differential Frame-Based Approach (DiffF), which incorporates explicit frame differences detected via computer vision techniques. We evaluate these methods using a basic self-curated dataset and an advanced benchmark adapted from prior work. Our results show that the DF approach achieves an accuracy of 70% to 80% in identifying user actions, with the extracted action sequences being re-playable though Robotic Process Automation. We find that while VLMs show potential, incorporating explicit UI changes can degrade performance, making the DF approach more reliable. This work represents the first application of VLMs for extracting user action sequences from desktop recordings, contributing new methods, benchmarks, and insights for future research.

Figures (3)

• Figure 1: Architectures of Direct Frame-Based Approach (left) and Differential Frame-Based Approach (right).
• Figure 2: An example output from Frame Difference Localizer. Left and right are the changed regions with red bounding boxes in the previous and current frames, respectively.
• Figure 3: An example output of Frame Difference Descriptor. It is derived from the identified UI changes in Figure \ref{['fig:comparator-example']}.