streetlevel-poles-yolo11n

YOLOv11n object detector for utility poles and towers in street-level imagery. Trained on the Street-level Poles & Towers dataset (curated by UC Berkeley, HOT-OSM YOLO experiments).

Classes

0: pole, 1: tower

Metrics (test split)

metric	value
mAP@50	0.332
mAP@50-95	0.120
precision	0.455
recall	0.345

10 epochs, imgsz=640. This is a sanity-check baseline, not a tuned model.

Try it in your browser

Interactive Gradio demo: https://huggingface.co/spaces/kshitijrajsharma/streetlevel-poles-detector

Inference (local)

from huggingface_hub import hf_hub_download
from ultralytics import YOLO

weights = hf_hub_download(
    repo_id="kshitijrajsharma/streetlevel-poles-yolo11n",
    filename="best.pt",
)
model = YOLO(weights)
results = model("street_image.jpg")
results[0].show()

Files

• best.pt — trained weights (Ultralytics format, ~5 MB).
• results.png, confusion_matrix.png, pr_curve.png — training diagnostics.
• metrics.json — machine-readable metrics summary.

License

CC-BY-SA-4.0.