Skip to content

kratos0718/apepdcl_internship

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

28 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

DTR Infrastructure Risk Assessment

APEPDCL AI/ML Internship — Problem Statement Submission


Problem

Field images of plinth-mounted distribution transformer installations need to be automatically assessed for safety risk based on the clearance height between the transformer base and the ground. The accepted minimum safe clearance is 6 feet.


Repository Structure

submission/
    Phase1_Problem_Understanding.md
    Phase2_Solution_Architecture.md
    Phase3_Approach_Assumptions_RiskMethodology.md
    Phase4_Evaluation_Strategy.md
    EdgeCase_Handling.md

prototype/
    dtr_risk_assessment.py    main inference pipeline
    train.py                  fine-tuning pipeline
    prepare_dataset.py        dataset splitting and YAML generation
    evaluate.py               metrics against ground truth
    app.py                    Streamlit dashboard
    requirements.txt
    sample_outputs/           annotated results on provided sample images

Setup

cd prototype
pip3 install -r requirements.txt

Running the pipeline

Single image:

python3 dtr_risk_assessment.py sample_outputs/sample_image_1.jpg

Folder of images:

python3 dtr_risk_assessment.py --folder /path/to/images --output results/

Running the dashboard

python3 -m streamlit run app.py

Opens at http://localhost:8501

Live demo: https://apepdclinternship-ekfck3gvxsxlymxfrd6yxi.streamlit.app/

The dashboard has two modes:

  • Sample images — select Image 1, Image 2, or run both and compare side by side. These are the two photographs provided in the problem statement PDF.
  • Upload your own — drag and drop any field photograph and get the assessment back.

Each result shows the annotated image, risk label, estimated clearance in feet, confidence score, scale calibration method used, and flags for vegetation encroachment or manual review.


Approach summary

Six-stage pipeline: preprocessing → object detection (YOLOv8) → ground plane estimation → scale calibration → clearance calculation → risk classification.

Risk levels: Safe (≥ 6 ft) / Risky (4–6 ft) / Highly Risky (< 4 ft)

Vegetation encroachment and barrier absence can escalate the base classification. Low-confidence outputs are flagged for manual review rather than guessing.

Full methodology, assumptions, risk logic, edge case handling, and evaluation strategy are documented in the submission/ folder.


Tarigoppula Sree Sai Abhinav

About

No description, website, or topics provided.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors

Languages