A combined CNN and LSH for fast plant species classification
Abdelkhalak Bahri, Karim El Moutaoikil, Imad Badi · 2019
The number of plants in the world is very important, to distinguish between them in a reasonable time can be problematic. Recent works has shown that convolutional neural network (CNN) models have great potential in the field of classification. The advantage of using this type of architecture, besides being robust, is that the network learns the characteristic vectors automatically thanks to the convolution layers. Due to his effectiveness, Locality Sensitive Hashing (LSH) becomes a popular method for large scale image classifcation. In this work, we propose a CNN-LSH method for fast plant species classification. Under the same conditions, our evaluation shows that our proposed method can reduce the time classification by 60 % compared to a CNN.