Flower search by image on mobile phone
Treepop Sunpetchniyom, Saowaluk Watanapa, Rungkarn Siricharoenchai · International Conference on New Trends in Information Science, Service Science and Data Mining · 2012
We propose an image search system for flowers on a mobile phone. Mobile phones have more limited resources than desktop computers in terms of CPU, RAM and data storage. The database that we created has 45 classes. We used 182 training images and 246 test images. We used an HSV histogram as a color feature. The accuracy rate using only color features was 44.86% with radius C=20. We use SURF as a shape feature. The accuracy rate using only shape-based featurewas 47.31% with SURF vectors S=25. We combine both color and shape features to achieved accuracy 61.61%.