Table 4 Implementation parameters and configuration for LeNet-5 model for criminal facial expression detection.
From: Criminal emotion detection framework using convolutional neural network for public safety
Parameter | Value |
|---|---|
Dataset | Emo135 (subset of 2000 images) |
Image size | 48 \(\times\) 48 \(\times\) 3 |
Normalization | Pixel values scaled to [0, 1] |
Train-Test split | 80% Training, 20% Testing |
Total classes | 135 (based on unique emotion labels) |
Model architecture | LeNet-5 |
1st Conv layer | 6 filters, 5\(\times\)5 kernel, ReLU, stride (1,1) |
1st MaxPooling | 2\(\times\)2 pool size, stride (2,2) |
2nd Conv layer | 16 filters, 5\(\times\)5 kernel, ReLU, stride (1,1) |
2nd MaxPooling | 2\(\times\)2 pool size, stride (2,2) |
Flatten Layer | Yes |
Dense layer 1 | 120 units, ReLU |
Dense layer 2 | 84 units, ReLU |
Output layer | 135 units (softmax activation) |
Optimizer | RMSprop |
Loss function | Categorical cross-entropy |
Batch size | 128 |
Epochs | 50 |
Validation split | 20% of data |
