Skip to content
Exploration of HOG, and Bag of Words features in the Caltech256 dataset.
Jupyter Notebook
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
imgs
.gitignore
README.md
challenges_of_image_classification.pdf
extraction-256_object_categories.ipynb
features-SIFT.ipynb
features.ipynb
image_classification.ipynb
models.ipynb

README.md

The Challenges of Image Classification

Before Convolutional Neural Networks came around, inferring what your images was a manual process of extracting features and piping them into a classifier. This project explores those classic techniques and compares three feature representations of images.

  • Compared histogram oriented gradients (HOG) and bag of visual words (BoW) using SIFT descriptors to raw pixel data using a LinearSVM.
  • Built BoW using OpenCV's implementation of SIFT and scikit-learn's KMeans.

Takeaways

The accuracy improved the more I abstracted the image, where SIFT created the largest accuracy, but I found that there was a problem with some classes not being guessed at all.

Tools

  • sklearn
  • OpenCV
  • skimage
  • LinearSVM
  • Mini-Batch KMeans
You can’t perform that action at this time.