Having worked on a real research paper, the university research paper now feels less daunting.
Over the past few months, he has been using this technology to help his team create a prototype that works with QCaption to describe images of crime scene evidence. QCaption is an AI tool that helps police officers annotate crime scene evidence and generate incident reports.
“While the original QCaption was good at describing evidence, its descriptions weren’t tailored for the task of crime scene logging,” he explains.
“The prototype I’m working on uses prompt engineering to first process images submitted by SPF officers before transmitting this curated data to the QCaption model. This ensures that QCaption generates more descriptive answers and helps SPF officers fill out crime logging forms faster, meaning that they can spend less time on paperwork and more time on investigations.”
Prompt engineering is the process of crafting an instruction that can be interpreted and comprehended by a Generative AI model.
Jiale helping Randall debug faulty code. (Photo: HTX)
To create this prototype, Randall received guidance from Wang Jiale, an engineer from the Q Team who also worked on QCaption.
“I am pretty new to Generative AI and was unfamiliar with the Large Multimodal Models used in this prototype. When Jiale noticed my struggles, he helped me understand the framework better, allowing me to add more exact prompts and fine-tune the prototype,” he shared.
Randall adds that he greatly appreciates the efforts that Jiale have made to check in on him frequently.
“During these sessions, Jiale was proactive and helped me solve any problems I faced before they snowballed into bigger issues. I also appreciate that he always took time to ensure I wasn’t overwhelmed by work,” he added.
Besides creating a prototype that has real world uses, Randall has also helped put together a research paper that will be presented at the 2024 International Conference on Information Fusion.
This research paper benchmarks QCaption’s video captioning capabilities against existing research, and positions QCaption as a game-changer in the artificial intelligence video captioning field.
Randall showing off the fruits of his labour. (Photo: HTX)
“Working on the research paper taught me valuable skills, like how to conduct literature reviews and experiments. These skills will really help me out in my final year of university, as I’ll need to write a research paper for my final year project. Having worked on a real research paper, the university research paper now feels less daunting!” Randall quipped.