tezvyn:

CNN features for image retrieval

Source: interviewadvanced

WHAT IT TESTS: transfer learning for features. OUTLINE: pass the image through a pretrained CNN and read activations from a late layer as a descriptor; deeper layers encode semantics, earlier layers encode texture.

WHAT IT TESTS: how to repurpose a classification network as a feature extractor. ANSWER OUTLINE: run the image through a network pretrained on a large dataset, drop the classification head, and take activations from a deep layer, often the last pooling or a fully connected layer, as a compact descriptor; pool and L2-normalize it. Deep layers capture high-level semantic content useful for retrieval, while earlier layers capture low-level texture. Compare descriptors by cosine or Euclidean distance.

Read the original → interview

Get five bites like this every day.

Tezvyn delivers a daily feed of 60-second tech bites with quizzes to lock in what you learn.

CNN features for image retrieval · Tezvyn