Project Info

reVix thumbnail

Team Name


Team Members

Mai and 3 other members with unpublished profiles.

Project Description

Revitalise. Reimagine. ReVix.

The Public Record of Victoria (PROV) houses images of immense historical values. However, they are heavily underutilised. While digitisation of these historic photographs is a great step forward, we, reVix, aims to bring these images to the public consciousness of the modern world.

Our goal
With reVix, we want to bring the historical images to the forefront and inject it to the public consciousness. For this objective, we have prototyped an app where users can view and easily generate a remixed version of a retrospective images. This way, we aim to give these old images a new coat of paint.

reVix will allow users to generate humorous captions for historical images recontextualising them. This will rejuvenate the users’ creativity giving a birth to creations that may hold enough wit to propel its circulation.

#meme #history #remix #revix #revitalise #reimagine #generative ai #image captioning #object detection #diffusers

Data Story

The PROV API contains 75 122 records of Photograph or Image. Unfortunately, out of 75 122 photo-graphics and images, only 35 774 are digitised. Furthermore, its existence is relatively unknown to the public. Each image carries a piece of Australian history, a snapshot of a moment of life and culture, all tucked away ready to be unveiled.

My personal favourite is H784 Welded fire dogs RMIT | Subject : IRON craft, a fascinating detail of a welded dog. I had found the figures adorable and thought what a pity it was as only a selected few were able to witness this snapshot of history, chronicled in a black and white. My curiously has led me seek out more snapshots of history using terms such as ‘craft’ and ‘giraffe’ only to be met with disappointment. The results had not been organised or filtered and I was overwhelmed with results irrelevant to my query.

This has led me to ponder ways to effectively filter and sort results. By sorting and organising results, it would allow users to discern images that they wished to view. One way was to implement a semantic search algorithm which required tagging, assigning descriptions, titles and other assortment of features of modern database.

For this reason, I had chosen to tackle Remix the Archives using the PROV API. What had started as a simple tagging project has expounded to something greater. We are now trying to close the gap between the old and the new, using remix to bring it to the public forefront. We are weaponising humour to hip the hop. Through reVix, we wish that searching others have fun surfing through the archives as I did.

Evidence of Work



Team DataSets

COCO dataset

Description of Use This dataset is used to train both pre-train models, BLIP image captioning model and YOLOS object detection model, which we used in our project.

Data Set

Public Record Office Victoria

Description of Use We use PROV API as a source to provide us images digitised and stored in Public Record Office Victoria's system.

Data Set

Challenge Entries

Tagging photographic images: showcasing the magnificent history of Victoria

How can we enable researchers to tag images of digitised records from photographic collections?

Go to Challenge | 6 teams have entered this challenge.

Remix the Archives using the PROV API

How might you use the PROV API to create a service that allows users to remix the archives for artistic endeavour?

Go to Challenge | 8 teams have entered this challenge.

Generative AI: Unleashing the Power of Open Data

Explore the potential of Generative AI in conjunction with Open Data to empower communities and foster positive social impact. This challenge invites participants to leverage Generative AI models to analyse and derive insights from Open Data sourced from government datasets. By combining the power of Generative AI with the wealth of Open Data available, participants can create innovative solutions that address real-world challenges and benefit communities.

Go to Challenge | 29 teams have entered this challenge.