perspectivLabs / CountingChallenge

0 stars 45 forks source link

KARTHIK PAI CODE #20

Closed KarthikPai14 closed 3 months ago

KarthikPai14 commented 3 months ago

Info

Name

KARTHIK K PAI

Python Version

PYTHON 3.12.

Description

AI

I utilize a zero shot segmentation method called MobileSAM by good for masking and counting. Use of MobileSAM ensures robust and highly efficient masking, with the use of lower resources compared to its parent SAM architecture. Since it is zero shot, it avoids the time curve and development cycle of creating and labelling datasets for the same. It also provides a option of further enhancement using image prompts.

Non_AI

We utilize a mix of canney edge detection and morphological transforms to enhance the edges in the image. This is followed by a straight forward contour detection to count the objects [in our case, nuts and bolts]. Utilizing this pipeline ensures a much more robust mechanism as each method in the pipeline complement each other and address each of their disadvantages.

Additional Comments

Loved the assignment as it was very similar to real world scenario problems-something which I have experience dealing with earlier.