Automated Video Analysis

By Dr. Emad Alghamdi
Dr. Alghamdi is an assistant professor of computational linguistics at King Abdulaziz University in Saudi Arabia. He is the director of the AILLA Lab, with research interests lie at the intersection of Applied Artificial Intelligence, Human-Computer Interactions (HCI), and Ethical AI. The project described in this post was awarded a Sage Concept Grant. Find him on Twitter @Emad_A_Alghamdi or LinkedIn: https://www.linkedin.com/in/emad-a-alghamdi/.


The Automated Video Analysis software (or AUVANA for short) is an open-source annotation tool for social scientists whose research involves analyzing and annotating videos. I developed the first version of AUVANA during my PhD. I remember searching for a tool that could help me annotate a large set of videos and analyze their content complexity, but I could not find a tool that is suitable for my research needs. So, I decided to develop my own and make it available to other researchers to use. Soon after I released the code, I received several emails from other researchers who wanted to use the tool for their research but found its functionalities limited. When I read a post on Twitter about Sage Concept Grant, I unhesitatingly applied, and my proposal thankfully got accepted. Rather than fixing the old tool, I decided to rebuild the tool from scratch. To help build the new tool, I reached out to some friends and developers from several online communities and explained to them the objectives of building AUVANA. Luckily, it did not take me long to convince three talented developers (Mishari, Abdulrahman, and Reem) to join me in this endeavor.

Before writing a single line of code, we reached out to potential users (researchers) and asked them how they currently analyze videos and what analyses and features their current tools lacked. We learned a lot from these interviews, and we were able to draw a roadmap for our tool.  Then, we spent a considerable amount of time deciding what technology stack we should go with and how to design the user interface, so it is user-friendly and scalable.  In the first version (1.00-Beta), we only included functionalities that are essential and yet do not take much time to implement. Another important engineering decision we made is to make our software functionalities extendable via plugins or addons (much like the R software). We envisioned that by developing an extendable platform other researchers and developers could join us to extend the tool's functionalities.

While there are many other alternatives, AUVANA is open-source software, and anyone can use or reuse its code in their projects. Secondly, it runs on the user's machine and stores their data in a local database, not in the cloud. Thirdly, it leverages cutting-edge AI technology to make video data annotation faster.

As we are now packaging our software and extensively testing it, it feels great that we could accomplish what we have done in a short period. It goes without saying, the path was not smooth, and we had moments of doubts and unclarity. Being all technical, we debated subtle things and sometimes rebuilt the same feature multiple times. We also implemented certain utilities, only to know later they do not work as we envisioned them. But we are glad now we have a working version, and we are excited to release it to the world soon at www.auvana.ai.

Have an idea? Apply for a Sage Concept Grant!

Sage’s Concept Grant program has been running since 2018 and aims to fund innovative software solutions that support research in the social sciences. We are seeking proposals for new technological solutions that support the adoption, development and application of established and emerging research methods, including quantitative, qualitative, mixed, and computational methods. We offer seed £2k grants to develop ideas and £15k grants for scaling prototypes. Explore previous winners. Apply for this year’s round and find out more here. Applications are open to anyone regardless of location or affiliation until 20th of June.


More Methodspace Posts about Data Analysis

 

Previous
Previous

Engaging with stakeholders: What can we learn from action researchers?

Next
Next

Using Secondary Data in Mixed Methods is More Straight-Forward Than You Think