Hernández-Luquin F, Escalante HJ, Villaseñor-Pineda L, Reyes-Meza V, Pérez-Espinosa H, Gutierrez-Serafín B. Dog emotion recognition from images in the wild: DEBIw dataset and first results. ACI '22: Proceedings of the Ninth International Conference on Animal-Computer Interaction. 2023; Article No. 5:1-13. https://doi.org/10.1145/3565995.3566041
Summary: This study introduces the DEBIw dataset, consisting of 15,599 internet-sourced images of dogs labeled for emotions such as aggression, anxiety, contentment, and fear. Multiple taggers manually labeled each image, and various image classification methods were tested for automatic emotion recognition, with an AutoML model achieving the best performance (macro average F1 score of 0.67). Notably, no pre-processing techniques such as segmentation or key point marking were applied to the raw images. The results show promise for developing non-invasive, easily retrainable systems for recognizing canine emotions in real-world scenarios.
Comments