| | |
| | | return get_network_delta_layer(net, net.n-1); |
| | | } |
| | | |
| | | float calculate_error_network(network net, float *truth) |
| | | { |
| | | float sum = 0; |
| | | float *delta = get_network_delta(net); |
| | | float *out = get_network_output(net); |
| | | int i; |
| | | for(i = 0; i < get_network_output_size(net)*net.batch; ++i){ |
| | | //if(i %get_network_output_size(net) == 0) printf("\n"); |
| | | //printf("%5.2f %5.2f, ", out[i], truth[i]); |
| | | //if(i == get_network_output_size(net)) printf("\n"); |
| | | delta[i] = truth[i] - out[i]; |
| | | //printf("%.10f, ", out[i]); |
| | | sum += delta[i]*delta[i]; |
| | | } |
| | | //printf("\n"); |
| | | return sum; |
| | | } |
| | | |
| | | int get_predicted_class_network(network net) |
| | | { |
| | | float *out = get_network_output(net); |
| | |
| | | return get_network_input_size_layer(net, 0); |
| | | } |
| | | |
| | | detection_layer *get_network_detection_layer(network net) |
| | | { |
| | | int i; |
| | | for(i = 0; i < net.n; ++i){ |
| | | if(net.types[i] == DETECTION){ |
| | | detection_layer *layer = (detection_layer *)net.layers[i]; |
| | | return layer; |
| | | } |
| | | } |
| | | return 0; |
| | | } |
| | | |
| | | image get_network_image_layer(network net, int i) |
| | | { |
| | | if(net.types[i] == CONVOLUTIONAL){ |