Compare baseline model predictions across datasets and weather conditions. Select a sample scene and explore predictions from three architectures against ground truth labels.
All views in a single horizontal strip: Input → Ground Truth → Model Predictions.