基于TensorFlow预训练模型快速、精准的图像分类器

A Fast & Accurate Image Classifier Based on the Pre-training Models of TensorFlow

图像分类是根据图像信息中所反映的不同特征,把不同类别的目标区分开来的图像处理方法,它利用计算机对图像进行定量分析,把图像或图像中的每个像元或区域划归为若干个类别中的某一种,以代替人的视觉判读.TensorFlow是Google基于DistBelief进行研发的新一代人工智能学习系统,TensorFlow是将复杂的数据结构传输至人工智能神经网络中进行分析和处理过程的系统.根据TensorFlow的预训练模型建立了一个低成本、快速、精准的图像分类器,根据实验结果,准确率达98%.

Image classification is an approach for processing images to distinguish the different types of target according to individual characteristics reflected in images.For the sake of replacing the human visual interpretation,it takes advantage of the computer version to carry on the quantitative analysis to images and classifies each pixel or regions in an image or the image into one of several categories.TensorFlow is a new machine-learning system developed by Google based on DistBelief.In TensorFlow,in order to analyze and process images,one needs to transfer the complex data structure to an artificial intelligence neural network.In this paper,we present a low-cost,fast and accurate image classifier on the basis of the pretraining models of TensorFlow.Experiment results show the classifier attains accuracy 98%.