---
license: agpl-3.0
pipeline_tag: object-detection
library_name: ultralytics
library_version: 8.3.86
tags:
  - ultralytics
  - object-detection
  - yolo
  - yolov12
base_model: Ultralytics/YOLOv12
---

## General description
All of Ultralytics' Yolo V12 models [model fined tuned](https://docs.ultralytics.com/modes/train/) for billboard detection using the [Billboard dataset](https://universe.roboflow.com/arslan-ongr8/billboard-xlvz1). 

This model was created with 100 epochs using CUDA 12.4 and Pytorch 2.6.0. 

# Best Metrics Comparison

| Model | Precision (Epoch) | Recall (Epoch) | mAP50 (Epoch) | mAP50-95 (Epoch) |
| ----- | ----------------- | -------------- | ------------- | ---------------- |
| YOLO_12n| 0.74539 (epoch: 56) | 0.67165 (epoch: 81) | 0.7079 (epoch: 70) | 0.42998 (epoch: 70) |
| YOLO_12s | 0.72631 (epoch: 33) | 0.67949 (epoch: 84) | 0.71221 (epoch: 70) | 0.43881 (epoch: 80) |
| YOLO_12m | 0.72904 (epoch: 53) | 0.67321 (epoch: 90) | 0.71937 (epoch: 74) | 0.43911 (epoch: 74) |
| YOLO_12l | 0.74444 (epoch: 94) | 0.6706 (epoch: 74) | 0.7145 (epoch: 79) | 0.4381 (epoch: 79) |
| YOLO_12x | 0.74258 (epoch: 80) | 0.68363 (epoch: 88) | 0.72273 (epoch: 76) | 0.45038 (epoch: 76) |


Further results can be found in [Results Folder](https://huggingface.co/maco018/billboard-detection-Yolo12/tree/main/results).

Created by [Mark Colley](https://m-colley.github.io/) - supported by [Zefwih](https://zefwih.com/)