Abstract
This work benchmarks three lightweight CNNs, MobileNetV2, EfficientNet-B0, and ResNet18, on a balanced AffectNet subset, evaluating accuracy, inference speed, and model size. Results show that lightweight CNNs provide competitive accuracy and efficiency, supporting their use in real-time applications.
| Original language | English |
|---|---|
| Title of host publication | Irish Machine Vision and Image Processing Conference 2025 |
| Publisher | Irish Pattern Recognition and Classification Society |
| Pages | 268 - 271 |
| Number of pages | 4 |
| ISBN (Electronic) | 978-0-9934207-9-5 |
| Publication status | Published (in print/issue) - 1 Sept 2025 |
| Event | IMVIP 2025 - Ulster University, Derry~Londonderry, Northern Ireland, Londonderry, United Kingdom Duration: 1 Sept 2025 → 3 Sept 2025 https://imvipconference.github.io/ |
Conference
| Conference | IMVIP 2025 |
|---|---|
| Country/Territory | United Kingdom |
| City | Londonderry |
| Period | 1/09/25 → 3/09/25 |
| Internet address |
Bibliographical note
Paper accepted and presented at IMVIP 2025UN SDGs
This output contributes to the following UN Sustainable Development Goals (SDGs)
-
SDG 3 Good Health and Well-being
Keywords
- imaging
- Machine Vision
- Emotion Recognition
- Deep Learning
- Lightweight CNN
Fingerprint
Dive into the research topics of 'Lightweight CNN Benchmarks for Facial Emotion Recognition in Pepper Robot Applications'. Together they form a unique fingerprint.Cite this
- APA
- Author
- BIBTEX
- Harvard
- Standard
- RIS
- Vancouver