SIFT, BoW Architecture and one-against-all Support Vector Machine.
Mohamed Issolah, Diane Lingrand, Fŕed́eric Precioso · 2013
Abstract. For this first participation to ImageClef Plant Identification, we build on the reference Bag-of-Word framework (BoW). We extract Points-of-Interest (PoI) using the SIFT detector in every image and de-scribe each local feature with the SIFT descriptor. The visual dictionary is built with a K-means algorithm of 100 clusters on the local features. Each image is then represented by its histogram onto the dictionary using hard-assignment strategy. We classify the images with as many binary one-against-all Support Vector Machines as the number of plant classes per organ types. Our aim is to evaluate for the plant identification task a classic baseline of multi-class image categorization. Our first results illustrate how difficult this task is and that a framework which has be-come a standard baseline for classifying general image datasets is not immediately relevant on Plant Identification data. 1 Our system Our system for ImageClef Plant Identification task [1] is built on the reference Bag-of-Word framework (BoW) and a set of binary Support Vector Machines. We use the XML metadata file provided with the images to extract information as Type of content or Background type. Let us now detail our settings. 1.1 Feature extraction and Image description We extract Points-of-Interest (PoI) using both the SIFT detector and the SIFT descriptor in each image. We extract about 1000 points in each image, with standard settings of Opencv C++ library: – Number of layers per Octave: 3 – The minimum threshold to consider a point as PoI: 0.04