Sentence examples for visualized feature from inspiring English sources

In this paper, a four-angle-star based visualized feature generation approach, FASVFG, is proposed to evaluate the distance between samples in a 5-class classification problem.

Table 1 Classification performance comparative results in two kinds of different target objects Visualizing feature extraction mode Classification performance Precision Recall Accuracy F1-Measure Lower layer visualizing feature Color 83.33 80 80.21 86 Texture 84.55 82 80.53 85 Higher layer visualizing feature paper proposed 91.75 84 82.15 87.

We respectively carry out the higher layer visualizing feature extraction and lower layer visualizing feature extraction in the target object images.

The results of classification performance in two kinds of target objects under the lower layer visualizing feature extracted and higher layer visualizing feature extracted this paper proposed are listed in Table 1.

Further, we will discuss the higher layer visualizing feature combining the lower layer visualizing features, such as color and texture, to find the most suitable methods to classify the small sample target objects in the complex sky background.

We adopt the color and texture of the lower layer visualizing feature to classify the different types of target object.

From Table 1, we can see the classification performance by using the higher layer visualizing feature extraction based on the SAE model is better than the traditional lower layer visualizing feature extraction, such as color and texture, which proved the accuracy of the higher layer visualizing feature extraction algorithm.

Based on the SAE model, the local higher layer visualizing feature (local feature) of small sample target object images can be extracted, and then, we can transform the local feature and the target object images to the CNN model; through a continuous convoluting and pooling process, the global higher layer visualizing feature (global feature) in the training set can be extracted.

During the higher layer visualizing feature learning processing, we adopt 400 number hidden layer units corresponding to 400 number self-learning features.

Because the sample numbers are enough, the classification precision under the SAE higher layer visualizing feature extraction based on non-transfer learning is slightly higher than transfer learning.

This paper proposed a kind of SAE higher layer visualizing feature extraction method for the small sample target objects in the sky.