The goal of assignment was to make real-world body part measurements using 2D images. The repository includes methods to measure shoulder distance, wrist-to-shoulder measurement, and waist approximation. For implementation details and other nitty-gritty associated with the project, its recommended to lookup the attached presentation named: Presentation.pdf.
To run code, change to src/ directory and type linux shell:
python code2.py -i1 <path to Image1> -i2 <path to Image2> -i3 <path to Image3> -a <Correction_mode>
2.4.13.6
True
to perform correction else False
waist
circumference. Mark waist ends on both the image.shoulder
length. Mark fall near neck both sides in first image and mark shoulder, both-sides in second image.wrist
- end of hands nearly in the next popped image. This image is with the subject holding a checkered board in hands. This helps measure shoulder distance. Check the image below.
Checkered board is special. Its helps in calibration of camera image world for 3D measurements. If you use any other chess-type board, measure the side length of unit square and change global ref_ht parameter in code2.py.
This image is with the subject spreading out his hands. This helps in wrist-to-shoulder measurement, and provide width of waist's projection.
This image is capturing side-view of subject. This provide thickness of waist and helps complete waist measurement.
Waist is modelled as an ellipse and measured analogous to finding perimeter of ellipse. Hence appromation is mentioned.
When code runs, you will be shown points selected by our heuristic as to-be shoulder/wrist. If suspicious/incorrect, you can explicitly select those points on image and pressing esc
key thereafter.