Gabor U-Net — Hyphae Segmentation

Upload any image (any format/size). It's reflect-padded and split into native-resolution 512x512 tiles (matching the training pipeline — no downscaling), each tile is run through the Gabor-attention U-Net (EfficientNet-B4 encoder), and predictions are stitched back together. Large images may take a while.

0 1
Tries native scale first; only retries at 1.5x/2x/2.5x if confidence is low — fixes photos where hyphae render too thin/thick vs. training tiles, at no extra cost to already-good uploads.
Suppresses detections that don't look hyphal (debris, nucleus rims, fold/coverslip edges). OFF by default because it is calibrated on too few slides: its shape priors were fitted on SPARSE, isolated filaments and they wrongly reject DENSE tangled hyphal mats, whose fragments are stubby and compact. On one such slide it discarded 25% of genuine hyphae. Turn it on only to clean up a slide you already believe is negative — never to rule Candida out.