# Which is the best algorithm for feature extraction in images?

Source: https://zilliz.com/ai-faq/which-is-the-best-algorithm-for-feature-extraction-in-images
Author: zilliz

The best algorithm for feature extraction depends on the application. Traditional methods like SIFT (Scale-Invariant Feature Transform) and HOG (Histogram of Oriented Gradients) are effective for tasks requiring handcrafted features, such as image matching or object detection in low-data scenarios. For deep learning applications, convolutional neural networks (CNNs) are the most effective, as they automatically learn hierarchical features from raw images. Pre-trained models like ResNet, EfficientNet, and Vision Transformers (ViTs) excel in feature extraction, particularly for large-scale datasets.