Publications

2024
Bitrate Ladder Construction Using Visual Information Fidelity
Krishna Srikar Durbha, Hassene Tmar, Cosmin Stejerean, Ioannis Katsavounidis, and Alan C. Bovik

  • We deploy features drawn from Visual Information Fidelity (VIF) (VIF features) extracted from uncompressed videos to predict the visual quality (VMAF) of compressed videos.
  • We present multiple VIF feature sets extracted from different scales and subbands of a video to tackle the problem of bitrate ladder construction.
  • We compare the performance of predicted bitrate ladders against a fixed bitrate ladder and a bitrate ladder obtained from exhaustive encoding using Bjontegaard delta metrics.
  • Our best-performing approach showed average BD-rate and BD-VMAF gains of 15.187% and 3.731 respectively, against Apple’s fixed bitrate ladder and BD-rate and BD-VMAF losses of 3.307% and 0.664 respectively, against a bitrate ladder obtained by exhaustive encoding.
[Paper] [Code]

2022
AutoML Models for Wireless Signals Classification and their effectiveness against Adversarial Attacks
Krishna Srikar Durbha and SaiDhiraj Amuru

  • We designed AutoML models of four different architectures namely Deep Residual Network (ResNet), Convolutional Long Short-Term Deep Neural Network (CLDNN), Convolutional Neural Network (CNN) and Recurrent Neural Network (RNN)
  • We studied and compared their performance of proprosed AutoML models with state-of-the-art on the task of wireless signal classification.
  • We also analyzed the vulnerability and effectiveness of the proposed AutoML models against transfer-based Projected Gradient Descent and Carlini-Wagner adversarial attacks
  • Our results show that AutoML models are a viable and solid candidate approach for the classification of wireless signals.
[Paper] [Code]