Open shijie404 opened 8 months ago
x[:, index] = conv_output[:, index]
Thank you for your attention on our work. First, the semantic compression is done in the file classification_crop.py, in which class CamExtractor_resnet is defined. The key idea is shown in the above code, we only choose the importance semantic features for transmission.
For the second question, since this repository is only for the proposed semantic compression method, the latency is not considered in this code. In addition, the transmission latency can be computed via (d*(1-o))/R, where d is the size of all semantic features, o is the compression, and R is the transmission rate, depending on the bandwidth and power.
Hello, I have two questions to ask. How are the formulas of semantic compression and latency threshold reflected in the code? Or is it not included in the project file? Thank you.