This repo implements a CBIR (content-based image retrieval) system
In this system, I implement several popular image features:
- color-based
- texture-based
- shape-based
- deep methods
Some features are not robust enough, and turn to feature fusion
The curse of dimensionality told that vectors in high dimension will sometime lose distance property
CBIR system retrieval images based on feature similarity
Robustness of system is evaluated by MMAP (mean MAP), the evaluation method is refer to here
- image AP : mean of precision at each hit
- class1 MAP = (class1.img[0].AP + class1.img[1].AP + ... + class1.img[M].AP) / M
- MMAP = (class1.MAP + class2.MAP + ... + classN.MAP) / N
My database contains 25 classes, each class with 20 images
Implementation of evaluation can found at evaluate.py
| Method | color | daisy | edge | gabor | HOG | vgg19 | resnet152 |
|---|---|---|---|---|---|---|---|
| Mean MAP (depth=10) | 0.614 | 0.468 | 0.301 | 0.346 | 0.450 | 0.914 | 0.944 |
Let me show some results of the system
Po-Chih Huang / @brianhuang1019