TCWeatherGen: Target-Consistent Weather Augmentation for Autonomous Driving

Diffusion editors can convert labeled clear-weather driving images into adverse-weather training data, but often deform task-critical targets such as traffic signs, lights, pedestrians, and vehicles, invalidating their annotations despite plausible weather appearance. Our analysis of FLUX.2 identifies three causes: aligned source attention, which carries a critical source constraint, weakens during denoising; background condition keys divert attention from target-specific support; and the global flow-matching loss gives small regions little spatial weight. We therefore introduce **TCWeatherGen**, which combines **SegRGBHeat** for sparse semantic anchoring, **Source–Condition LoRA (SC-LoRA)** for object-aware pathway adaptation, and **Content-Aware Token Pruning (CATP)** for suppressing uninformative condition tokens. On ACDC weather transfer, our fine-tuned 4B student improves target ROI SSIM from $0.560$ to $0.727$ and FID from $83.6$ to $57.1$ against matched FLUX.2-32B-dev pseudo-targets. CATP reduces peak training and inference memory by 10 and 3 GB and inference time by $35.8\%$. TCWeatherGen augmentation further improves downstream mIoU and mAP${}_{50}$ from $0.484$ and $0.239$ to $0.598$ and $0.339$.