Analyzing sedentary behavior in life-logging images

Abstract

We describe a study that aims to understand physical activity and sedentary behavior in free-living settings. We employed a wearable camera to record 3 to 5 days of imaging data with 40 participants, resulting in over 360,000 images. These images were then fully annotated by experienced staff with a rigorous coding protocol. We designed a deep learning based classifier in which we adapted a model that was originally trained for ImageNet [1]. We then added a spatio-temporal pyramid to our deep learning based classifier. Our results show our proposed method performs better than the state-of-the-art visual classification methods on our dataset. For most of the labels our system achieves more than 90% average accuracy across different individuals for frequent labels and more than 80% average accuracy for rare labels.

Usefulness of Wearable Cameras as a Tool to Enhance Chronic Disease Self-Management: Scoping Review (Preprint)

“…Eight had between 20 -50 participants [8,9,13,20,24,[29][30][31][32][33]36,37] and only three studies (13%) [26][27][28] had more than 50 participants (maximum = 84). Participants were predominantly healthy adults (n= 15, 65%) [8,[10][11][12][13]20,22,23,[25][26][27][28]33,35,36], with some studies including athletes (n = 4, 17%) [24,[29][30][31] and school children (n = 2, 9%) [9,34]. One study used a sample of older women to examine physical activity and sedentary time [32].…”

“…Studies including data analysis were predominantly feasibility or pilot studies (n = 13, 54%) [9][10][11]16,20,[22][23][24][25][26][27][28][29], followed by validation studies (n = 4, 17%) [8,28,30,31] and methodological studies (n = 5, 21%) [32][33][34][35][36]. There was one randomised controlled trial [37] conducted with acquired brain injury patients where camera images formed part of a health intervention.…”

“…There was one randomised controlled trial [37] conducted with acquired brain injury patients where camera images formed part of a health intervention. The majority of the studies were conducted in the United States of America (n = 7, 29%) [22,25,27,29,31,32,36], the United Kingdom (n = 6, 25%) [9][10][11][12][13][14], and Ireland (n = 4, 13%) [24,30,33,35]. Three studies [23,26,28] were international multicentre studies, while the remaining studies were from New Zealand [8,20,34] and Spain [37].…”

“…Studies classified as validation [8,28,30,31] compared the validity of wearable cameras either against, or in combination with other measurement approaches, including self-reported diaries, questionnaires, as well as objective measurement techniques such as accelerometers and doubly labelled water. For the methodological papers [32][33][34][35][36], outcomes focused on testing software or analytic approaches to classify and analyse generated images.…”

“…In 21 studies (81%) participants wore a camera for between three to seven days; in one study, participants wore the cameras for seven weeks [37]. Studies classified as primary or secondary data analyses used wearable cameras to measure, observe or validate specific behaviours related to dietary intake (n = 10, 38%) [8][9][10][11]20,[22][23][24][25]39], physical activity (n =5, 17%) [26,27,30,32,40], a diverse range of daily activities such as travel, work and exposure to food marketing (n = 5, 17%) [12,[33][34][35]37], sedentary behaviour (n = 4, 15%) [13,29,31,36] and travel behaviours such as walking, cycling or motorised transport (n = 2, 8%) [14,28].…”