DeepSense Forum
Dear admin,
Hi, I am Irfan from Sejong University. I really interested to your works in blockage prediction task. I tried to reproduce your work by using vision based future blockage prediction dataset. However i struggled to understand how to predict future blockage when n>1. Can you give me your snippet of your code ? Thank you
Best regards,
Irfan
Dear Irfan,
Thank you for your interest in our work on blockage prediction. This work aims to enable the base station to predict whether or not the stationary user will get blocked within a window of r^{\prime} (or n, as you have mentioned) future instances. In other words, given a sequence of input images, the model predicts whether there will be a blockage in the future n instances.
For further clarity on how the future link status is computed, please refer to Eq. 4 in our paper ( https://ieeexplore.ieee.org/abstract/document/10008524 ). In summary, if there is any blockage during these time slots, the blockage status for the 'n' time slots (future-n) is considered blocked. This link status ('1' or '0') is then used as the ground-truth label during training and evaluation.
We also recommend that you review the evaluation code on the GitHub repository ( https://github.com/gourangc/Vision_Aided_Blockage_prediction ). If you have any further questions or require additional assistance, please let us know, and we will be happy to help.
Best regards,
Team DeepSense
