Ocean-router: Marine Image Routing Classifiers

Two-stage lightweight classifiers used to dynamically route marine images to specialized detectors based on image modality and content.

Model Files & Tasks

File	Task	Architecture	Input/Output
cls_bio_sonar/best.pt	Sonar vs. Biological routing	YOLOv11-cls	Image → [sonar_prob, bio_prob]
fish_coral_cls/best.pt	Fish vs. Coral routing	YOLOv5	Image → [fish_prob, coral_prob]

Usage

1. Sonar/Biological Router

from ultralytics import YOLO
router = YOLO("cls_bio_sonar/best.pt")
results = router.predict("input.jpg")

2. Fish/Coral Classifier (YOLOv5)

Requires the official YOLOv5 repo

import torch
model = torch.hub.load("ultralytics/yolov5", "custom", path="fish_coral_cls/best.pt", force_reload=True)
results = model("input.jpg")

cls_bio_sonar decides if input is sonar or biological. If biological, fish_coral_cls routes to the appropriate species detector.