Building Large-Scale Image Feature Extraction with BigDL at JD .com Artificial Intelligence Image Feature Extraction

Let’s examine the architecture of the image feature extraction application. As the background of many images can be very complex, and the main object in the image is often small, the main object needs to be separated from the picture’s background for correct feature extraction. Naturally, the framework of image feature extraction can be divided into two steps. First, the object detection algorithm is used to detect the main object, and then the feature extraction algorithm is used to extract the features of the identified object. Here, we use the Single Shot MultiBox Detector* (SSD)2 for object detection, and the DeepBit* model3 for feature extraction.