Footwear-Classification

Footwear Type and Color Identification

Dependencies

• python -> 2.7
• keras -> 2.0
• tensorflow -> 1.0
• sklearn -> 0.18

Report

Report for the project is saved as report.docx

Executable code

Executable code is in Prediction folder. The make_predicitions notebook can be referred on how predictions are being made. Instructions to execute the code are mentioned in readme.md in the precition folder and also in the python(make_predictions.py) file. Refer to the comments for step be step instructions and download the necessary data and classifiers(pickled) files.

Footwear Type Folder

Contains jupyter notebooks for:

• saving images
• Fine Tuning using CNN
• Extracting features
• Applying PCA transformation
• Training type_classifier using Support Vector Machines

Note: The files resulting as output on running these notebooks are not included as the prediction code can be executed without them. Also their size are very large.

Footwear Color Folder:

Notebooks for training color_classifier

• All the notebooks are somewhat similar to the notebooks in Footwear Type folder with few changes
• Validation and fine tuning notebooks are not present as fine tuning fails on small images and validation set was not built due to size constraints

Note: Comments in the notebooks can be referred to learn more about how classifiers were trained