https://github.com/l0sg/shoes_classification
Simple shoes classification with ResNet50 in Keras for Deepest Hackaton Challenge
https://github.com/l0sg/shoes_classification
fine-tuning image-classification keras resnet-50 scikit-learn transfer-learning
Last synced: about 1 month ago
JSON representation
Simple shoes classification with ResNet50 in Keras for Deepest Hackaton Challenge
- Host: GitHub
- URL: https://github.com/l0sg/shoes_classification
- Owner: L0SG
- Created: 2017-05-28T06:12:56.000Z (over 8 years ago)
- Default Branch: master
- Last Pushed: 2017-10-18T08:02:55.000Z (almost 8 years ago)
- Last Synced: 2025-07-11T07:59:36.154Z (3 months ago)
- Topics: fine-tuning, image-classification, keras, resnet-50, scikit-learn, transfer-learning
- Language: Python
- Homepage:
- Size: 1.74 MB
- Stars: 7
- Watchers: 2
- Forks: 3
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# Shoes classification
Simple shoes classification with ResNet50 in Keras for Deepest Hackaton Challenge.
Achieves ~77% accuracy in validation dataset.
It uses pre-trained ResNet50 for transfer learning.
It unfreezes the last 2 redisual blocks for fine-tuning.
## How to use
Requires Keras and Scikit-learn
Run train.py to fine-tune ResNet50 with the dataset.
You could use any arbitaray image dataset, by appropriately constructing your images with subfolders and modifying the class names in the code
Run inferencce.py to validate the model with test dataset.
## Download Shoes Dataset
You can download the dataset here
https://drive.google.com/open?id=0B05wqMLeCbG-blZpVTlEZWQ3NWs
It contains 14 classes in jpg image format with variable sizes (crawled from Google & Naver image search)
Unzip and specify the location of the dataset at datapath parameter in load_imgs function