Mini-project: Foot size estimation

https://github.com/dang-nh/FootSize-Estimation

A demo website using Streamlit with the following functionalities:

• upload images

• foot and paper segmentation: possibly uses image processing techniques or DL-based approaches

• visualize intermediate results: segments of foot, paper, paper's corners...

• output width, height and corresponding size

How to Evaluate?

• Clone the repo: git clone https://github.com/dang-nh/FootSize-Estimation.git
• Download folder image_classified from this repo or from Goolge Drive: https://bit.ly/3I7Qgyd
• cd to the cloned directory
• Install required packages: pip install -r requirements.txt
• run the command: streamlit run streamlit_demo.py

Working approach

• Classify type of image by SVM
• Remove noise from original image using Median Blur
• If image background have same color with skin, using Gamma Correction to improve contrast and tuning upper bound and lower bound of HSV to get better result
• Run K-means clustering on preprocessed image for color based segmentation
• Detect the edges in clustered image.
• Find contours in Edge Detection output and filter the largest contour.
• Generate the bounding rectangle from the contour and rotate to extract paper
• Then drop 10% of the paper image to get the foot contour.
• Create the foot's bounding Box to get the height/width of Foot.
• Calculate the foot size base on the ratio between the paper and foot. https://www.dienmayxanh.com/kinh-nghiem-hay/cach-xac-dinh-size-giay-cho-nam-gioi-don-gian-de-1357862

Limitations

• If floor color is white, then it will difficult to segment the paper but if using K-means with num_cluster=3 it will be easier.
• Feet should not go out of the paper.
• If the background have many patterns and have some white lines, then it will be difficult to segment the paper.
• The accuracy of the algorithm is not very good. It is about approx. 65% accurate.

Group members

• Nguyen Hoang Dang 20194423
• Do Quoc An 20194414
• Ha Vu Thanh Dat 20194424
• Le Hai Son 20194449