StereoGenBench: A Synthetic Multi-Camera Benchmark for Stereo Generation under Controlled Baseline Regimes
Researchers have introduced StereoGenBench, a new synthetic dataset created in Unreal Engine for evaluating stereo generation and view synthesis. This benchmark provides meticulously controlled camera baselines, intrinsics, scene depth, and motion data across multiple calibrated views. It aims to enable precise measurement of how sensitive stereo generation models are to varying baseline configurations and ensure consistency across different camera setups. AI
IMPACT Provides a controlled environment for developing and evaluating stereo vision models, potentially improving 3D reconstruction and augmented reality applications.