Content-Based Image Retrieval System for Dermal Images using Transfer Learning

Mohamed S. Elbially • Sara Abdeldayem • Dalia Bassiouny • Inas Yassine

09:05 - 09:55 | Thursday 16 February 2017 | Ballroom D


Content-based image retrieval (CBIR) is an emerging need for facilitating decision systems in the medical contexts. In this study, we investigate a CBIR platform for dermal lesion retrieval. The feature set used in this study is extracted based on the Fined Tuned AlexNet Convolution Neural Network. Four Different similarity measures have been used to perform the retrieval task. The proposed system achieved a precision of 98% and 97% using correlation and Euclidean similarity measures respectively.