Fully convolutional network for recognition of five emotions ('neutral', 'happy', 'sad', 'surprise', 'anger').
For the metrics evaluation, the validation part of the AffectNet dataset is used. A subset with only the images containing five aforementioned emotions is chosen. The total amount of the images used in validation is 2,500.
| Input Image | Result |
|---|---|
Happiness |
| Metric | Value |
|---|---|
| Input face orientation | Frontal |
| Rotation in-plane | ±15˚ |
| Rotation out-of-plane | Yaw: ±15˚ / Pitch: ±15˚ |
| Min object width | 64 pixels |
| GFlops | 0.126 |
| MParams | 2.483 |
| Source framework | Caffe |
| Metric | Value |
|---|---|
| Accuracy | 70.20% |
[*] Other names and brands may be claimed as the property of others.