Vision by Black
We are a team of computer scientists passionate about computer vision and machine learning. Over the course of this weekend we have been exposed to immense repositories of public data, of which large portion lies in visual image formats.
Through talking to a number of mentors and problem owners, it has become apparent that a common problem faced by CSIRO, TERN, ALA and similar scientific data collectors is the tagging, deep linking, sorting and general classification of image meta data. Adding this type of information to images is a hugely labourious task and is still accomplished manually. Whilst this human resource isn't much of a problem with small databases, it consumes large amounts of (mainly) volunteer time when dealing with databases on the scale of tens-of-thousands of images.
To solve this problem we have created a web service to automate the classification of image meta data, using various trained models and services to help machines recognise valuable information that us humans obtain with just a glance. In doing so, our system can tag images in a meaningful way, giving people easy access to a range of images on various topics regardless of their library location.
The service is now live at:
for a demo.