Uses the 2D geometric shapes dataset [1].
Achieves approx. 95% - 97% accuracy.
Utilizes a CNN implemented using pytorch for classification.
Written out of personal interest
Run shape_preprocessor with path to image folder as argument. This creates a pickle of the processed images within the source folder. Then, just run classifier.py.
- pytorch
- numpy
- cv2
- sklearn
[1] Anas, EL KORCHI (2020), “2D geometric shapes dataset ”, Mendeley Data, V1, doi: 10.17632/wzr2yv7r53.1 http://dx.doi.org/10.17632/wzr2yv7r53.1 https://data.mendeley.com/datasets/wzr2yv7r53/