Table 2.
Overview of food image databases for autonomic recognition studies.
| Databases | Coverage of affective space | Recording methods | Cuisines | Availability of normative (and demographic) data | Remarks |
|---|---|---|---|---|---|
| PFID (Chen et al., 2009) | A small part of valence and arousal space | Not standardized | Mainly Western | Not available | • High resolution (2592 × 1944 pixels) |
| • Collected by the authors | |||||
| • No fixed background | |||||
| NU FOOD (Takahashi et al., 2017) | Mainly positive valence | Standardized | Some Asian, Some Western | Not available | • Only 10 different cuisines (six Asian and four Western cuisines) |
| • No resolution specified | |||||
| ChineseFoodNet (Chen et al., 2017) | Mainly positive valence | Not standardized | Only Chinese | Not available | • Variable resolution |
| • Collected from Internet (185,628 images) | |||||
| • No fixed background | |||||
| UNICT Food Dataset 889 (Farinella et al., 2015) | Mainly positive valence | Not standardized | Italian, English, Thailand, Indian, Japanese etc. | Not available | • Variable resolution |
| • Collected with smartphones (3,583 images) | |||||
| • No fixed background | |||||
| UEC-Food 100 (Matsuda et al., 2012) UEC-Food 256 (Kawano and Yanai, 2015) | Mainly positive valence | Not standardized | France, Italy, United States, China, Thailand, Vietnam, Japan, Indonesia, etc. | Not available | • Variable resolution |
| • Collected from Internet | |||||
| • No fixed background | |||||
| UPMCFOOD-101 (Wang et al., 2015) ETHZFOOD-101 (Bossard et al., 2014) | Mainly positive valence | Not standardized | More than 101 international food categories | Not available | • Variable resolution |
| • Collected from Internet | |||||
| • No fixed background | |||||
| VIREO-172 (Chen and Ngo, 2016) | Mainly positive valence | Not standardized | Only Chinese | Not available | • Variable resolution |
| • Collected from Internet (110,241 images) | |||||
| • No fixed background | |||||