Reproducible deep-learning image-classification pipeline for industrial soft sensing. Originally built for sludge-cake quality monitoring at DC Water Blue Plains AWWTP. Six architectures (FastViT, EfficientNet, MobileNet, EfficientFormerV2, DeepTEN-ResNet, sparse-AE CNN) compared with multi-seed statistics on Modal cloud GPUs.
computer-vision deep-learning modal pytorch environmental-engineering image-classification transfer-learning mobilenet cloud-gpu wastewater-treatment efficientnet soft-sensor timm fastvit dc-water
-
Updated
May 10, 2026 - Python