Reproduce “CIFAR-10 Image Classification Using Feature Ensembles” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Felipe O. Giuste, Juan C. Vizcarra
The authors combine five image-feature sources—transfer-learned CNNs, HOG, pixel intensities, and CIFAR-VGG features—to classify CIFAR-10 images. They reduce the combined features to 1,000 principal components with PCA, then train a fully connected neural network. The system reaches 94.6% accuracy, versus 93.43% for CIFAR-VGG; this small margin is reported only on CIFAR-10, with test-set validation during training.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.