Poles_LeanedStraight/ObjectDetection/README.md

---
license: cc-by-4.0
---

# Object Detection — Leaned vs Straight Poles

- This dataset contains real-world images of electric utility poles labeled as either **"Leaned_Pole"** or **"Straight_Pole"**, It represents the initial stage in a broader effort to identify various defects in electric poles using machine learning.
- This module trains and evaluates a YOLOv12 model to detect **Leaned_Pole** vs **Straight_Pole** in field images.

## Dataset Details

The dataset is organized into pre-split folders:
```
Poles_LeanedStraight/ObjectDetection/
├── train/
│ ├── images/
│ └── labels/
├── valid/
│ ├── images/
│ └── labels/
├── test/
│ ├── images/
│ └── labels/
└── data.yaml
```

### Dataset Description

The dataset comprises **1804 real-world annotated images** of electric poles collected across **Visakhapatnam, Eluru, and Kakinada** districts in Andhra Pradesh, India. These images were captured by utility linemen using mobile phone cameras during field inspections.

Each image has been annotated in **YOLOv12 format** with one of two labels:
- `Leaned_Pole`
- `Straight_Pole`

- **Curated by:** Sampath Balaji & team, at APEPDCL
- **Funded by [optional]:** Eastern Power Distribution Company of Andhra Pradesh Limited (APEPDCL)
- **Language(s) (NLP):** Not Applicable
- **License:** CC BY 4.0

### Dataset Sources

- **GitHub Repository (code):** [Direct Link](https://github.com/EPDCL/Electrical-Lines-Defect-Detection/tree/main/Pole_LeanedStraight_Defect/ObjectDetection)


## Uses

### Direct Use

This dataset is intended to train, evaluate, and benchmark machine learning models for infrastructure monitoring, specifically for identifying leaned and straight poles in real-world, diverse conditions.

### Out-of-Scope Use

- Surveillance or facial recognition tasks
- Biometric identification
- Any use that violates personal rights or local laws

## Dataset Structure

- **Total images:** 1804
- **Split:**
  - Train: 1444
  - Validation: 181
  - Test: 179
- **Labels:** `Leaned_Pole`, `Straight_Pole`
- **Format:** JPEG images with annotations in YOLOv12 format
- **Augmentations:** None applied

## Dataset Creation

### Curation Rationale

This dataset was created to build an ML pipeline for real-time defect detection in utility poles, starting with identifying leaned poles as a foundational task. The aim is to build an electrical line quality monitoring system to revolutionize maintenance workflows.

### Source Data

#### Data Collection and Processing

Images were collected using mobile cameras by field linemen in Andhra Pradesh. All annotations were manually reviewed and labeled in YOLOv12 format.

#### Image Capture Protocol

The images were captured on the ground by APEPDCL linemen following a field protocol designed to reduce bias and ensure consistency:

- Linemen were instructed to **hold the phone upright and parallel to the pole** to avoid tilt or skew that could distort the pole's perceived angle.
- Images were to include both the **top and bottom of the pole**, as full visibility is essential for lean classification.
- **No filters or enhancements** were applied during capture.

Despite this protocol, many images were captured with a **tilted horizon**, making it difficult to determine if a pole was leaned or not. Tools like **Google Photos** and **Adobe Lightroom** were used to **manually correct tilt** by straightening the horizon.

To prevent this in future data, the team identified an open-source camera app, **[Open Camera](https://opencamera.org.uk/)**, which uses the phone’s **gyroscope sensor to auto-straighten images**. This app is now recommended for field use to ensure more consistent alignment.

#### Observations and Future Protocol Improvements

During annotation, a key insight emerged:

- When images were taken **too close to the pole**, **perspective distortion** made it harder to determine pole alignment.
- Going forward, linemen will be instructed to **stand approximately one pole height away** to reduce distortion and improve labeling accuracy.

#### Who are the source data producers?

Electric linemen from APEPDCL working in Visakhapatnam, Eluru, and Kakinada districts of Andhra Pradesh, India.

### Annotations

#### Annotation process

Manual bounding box annotations were created using image annotation tools and saved in YOLOv12 format. The team worked to maintain consistency, though **some labeling errors may persist due to the manual nature** of the task.

#### Who are the annotators?

The APEPDCL project team, under the direction of the team lead.

#### Personal and Sensitive Information

No personal or sensitive data was intentionally included. However, some images may contain **people or faces captured incidentally** in public areas, as the data was collected in outdoor settings.

## Bias, Risks, and Limitations

- **Geographic bias:** All data comes from Andhra Pradesh, India — results may not generalize to other regions or environments.
- **Scene/perspective bias:** Variation in camera angles and distances may affect model training.
- **Privacy risk:** Images taken in public areas may contain identifiable people.

### Recommendations

- Apply **image normalization or augmentation** techniques to improve generalization.
- Account for **perspective and camera angle** effects during training and evaluation.
- Use ethically and responsibly, especially in systems that affect public infrastructure.

## Metrics on dataset:
### Validation Results (conf 0.35)
| Class          | Precision | Recall    | mAP\@0.5  | mAP\@0.5–0.95 |
| -------------- | --------- | --------- | --------- | ------------- |
| Leaned\_Pole   | 0.894     | 0.894     | 0.963     | 0.730         |
| Straight\_Pole | 0.914     | 0.875     | 0.934     | 0.572         |
| **Overall**    | **0.904** | **0.884** | **0.949** | **0.651**     |

### Test Results (conf 0.35)

| Class          | Precision | Recall    | mAP\@0.5  | mAP\@0.5–0.95 |
| -------------- | --------- | --------- | --------- | ------------- |
| Leaned\_Pole   | 0.911     | 0.807     | 0.928     | 0.734         |
| Straight\_Pole | 0.922     | 0.879     | 0.968     | 0.630         |
| **Overall**    | **0.917** | **0.843** | **0.948** | **0.682**     |

## Sample predictions
<p align="center">
  <img src="https://raw.githubusercontent.com/sampath-balaji/Electrical-Lines-Defect-Detection/refs/heads/main/Pole_LeanedStraight_Defect/ObjectDetection/assets/output.jpeg" width="600"/>
</p>


## More Information [optional]

Follow-up datasets in this series will include other electrical line and pole defects as part of APEPDCL's Line Quality monitoring project.

## Dataset Card Authors

- [Sampath Balaji](https://github.com/sampath-balaji)

## 📄 License
- [Code](https://github.com/EPDCL/Electrical-Lines-Defect-Detection/tree/main/Pole_LeanedStraight_Defect/ObjectDetection): MIT License
- [Dataset](https://huggingface.co/datasets/EPDCL/Electrical-Lines-Defect-Detection/tree/main/Poles_LeanedStraight/ObjectDetection): CC BY 4.0