Skip to content

Latest commit

 

History

History
38 lines (37 loc) · 3.65 KB

README.md

File metadata and controls

38 lines (37 loc) · 3.65 KB

EVRAZ_CV_HAKATHON

This repository contains training and inference scripts for the EVRAS track 2 (computer vision) solution by Ammar Ali.

Note

Don't forget to download the data

Installation and Running (Human Detection)

Option 1

  • pip install -r requirements.txt
    

  • Download weights from here
  • python run.py --weights weights_path --data images_path --output output-dir --sub_sample sample_submission.json --imgsz image_size
    

Option 2

  • docker build .
    

Visualization

Ground Truth

Alt Text

Predictions

Alt Text

Installation and Running (Danger Detection)

Option 1

  • pip install -r requirements.txt
    

  • Download weights from here
  • python run.py --img_path orig1.jpg --weights tf_efficientnet_b1_ns_fold0_best.pth
    

Option 2

  • docker build .
    

Visualization

Original Image VS Heat Map