Documentation of:
45.72 x 45.72 cm
Laser prints mounted on paper
15 x 15 cm 
Photobook, 82 pages
Frustrated with my growing library of unsorted digital photos, I turned to a computer vision software— a software that can quickly classify images into predefined categories up to human standards—for help. "Things: 625 images of _____" picks up on some of the labels my photos were automatically sorted into through the software’s convolutional neural network (CNN)—a class of deep neural networks used to analyse images using raw pixel data as input. I replicated this process by punching out a portion of each image that reflects one of the features the CNN could have extracted and processed to classify them in this manner. In doing so, I reflect on what I photograph, why I photograph and the value of a photograph.

In this project, I picked up the top keywords the CNN generated based on my library. From each keyword, I randomly selected 625 images from the whole collection to print out and punch 1 cm x 1 cm square holes. For most images, it was easy to identify the characteristics that inform the CNN how to classify it. For example, in “Birthday”, it would be birthday cakes or candles. However, in some images, it was hard to tell what made the CNN classify it that way. To humans, it would be considered an incorrect classification. In the photobook that accompanies the prints, I compiled some of these incorrectly categorised images. 
Things: 625 images of “sunset”
Things: 625 images of “park”
Things: 625 images of “birthday”
Shooting Home Youth Awards Class of 2020 Exhibition, Objectifs, Singapore (2021)