search

perspectivLabs / CountingChallenge

0 stars 45 forks source link

mayuresh satam solution #21

Open Mayuresh1611 opened 3 months ago

Mayuresh1611 commented 3 months ago

Info

Name

Mayuresh Satam

Python Version

Python 3.12.4

Description

AI

I used yolov8 model and trained it over masked images. Object masks for 5 images were created to train the model,

Non_AI

  1. The image is first grayscaled and then adaptive thresholding is applied to remove the shadows as well as the vignette.
  2. After detecting edges using Canny edge detection, contours are extracted,
  3. These contours are classified into screws and nuts based on the contour edges, (the hole of the nut is considered for classification )
  4. Screws and nuts are filtered out from the noise produced by respective contours.

Additional Comments

The methods presented may be biased to the dataset provided, and might fail to detect objects outside of these dataset.