Event level emotion recognition works reliably on wild photos
Event-Level Emotion Recognition in the Wild Using Deep Facial Expression Analysis
Computer Vision and Pattern Recognition
Summary
Recognizing emotions from faces in crowded, real-life photos is hard because of different lighting, poses, and overlapping faces. The authors developed a system that first finds faces, then uses a smart neural network trained on expression data to identify emotions on each face. Finally, it combines all these to figure out the overall mood at public events from photos. They tested various neural nets and found one called EfficientNet-B2 works best for photos taken in the wild. Their method gave consistent results on many real event photos, showing it can tell the general emotional atmosphere of an event.
What this means in practice
- •For event organizers: Assess and monitor crowd emotions efficiently from event photographs to improve event management and attendee experience.
- •For security monitoring teams: Detect collective emotional states at public gatherings to better understand crowd mood and potential risks.
Authors
Aleksandr Semerikov, Pakizar Shamoi
Abstract
Facial emotion recognition (FER) in real-world environments remains challenging due to unconstrained imaging conditions, including multiple faces, occlusions, pose variations, and complex lighting. Most existing studies focus on individual facial emotion classification and do not address the analysis of collective emotional states at the event level. This paper proposes an end-to-end pipeline for event-level emotion recognition from photographs. The approach detects faces in each image, classifies facial expressions using a deep convolutional neural network, and aggregates face-level emotion probabilities to estimate the overall emotional distribution of a public event. A comparative evaluation of several CNN architectures on the FER- 2013 and RAF-DB datasets demonstrates that transfer learning with EfficientNet-B2 trained on RAF-DB is more suitable for real-world RGB data. The proposed method is evaluated on a real-world event dataset containing 1658 images. Experimental results show stable emotion distributions across event subsets, confirming the effectiveness of event-level aggregation for emotion analysis in the wild.