Parameter and Data-Efficient Spectral StyleDCGAN

We present a simple, highly parameter, and data-efficient adversarial network for unconditional face generation. Our method: Spectral Style-DCGAN or SSD utilizes only 6.574 million parameters and 4739 dog faces from the Animal Faces HQ (AFHQ) dataset as training samples while preserving fidelity at low resolutions up to 64x64. Code available at https://github.com/Aryan-Garg/StyleDCGAN.

Analyzing the Robustness of PECNet

Comprehensive robustness analysis of PECNet, a pedestrian trajectory prediction system for autonomous vehicles. A novel metric is introduced for dataset analysis and classification. Synthetic data augmentation techniques ranging from Newtonian mechanics to Deep Reinforcement Learning based simulations are used to improve and test the system. An improvement of 9.5% over state-of-the-art results is seen on the FDE while compromising ADE. We introduce novel architectural changes using SIRENs for higher precision results to validate our robustness hypotheses. Additionally, we diagrammatically propose a novel multi-modal system for the same task.