In this research, we are studying on image-based identification of trees species that we can see everywhere. In our previous study, we showed that convolutional neural network (CNN) can recognize tree species by using a region of interest (ROI) image of bark. However, the bark region is manually extracted from a natural bark image. This paper solves this problem using semantic segmentation, and proposes an automatic tree species identification from natural bark image. The proposed method was evaluated with the bark image dataset collected independently. We confirmed the effectiveness of the proposed method.
This paper proposes a convolutional neural network (CNN)-based tree species identification method from bark image. The proposed method uses the well-known CNN model. The difficulty of our problem is to use a special tree image in which a colorful tag is stick on the bark. The tag is irrelevant to the species. In order to recognize with CNN, it is necessary to extract a region (ROI) excluding the tag. Thus, this paper proposes a ROI extraction method. Extracted ROI is fed to CNN. We evaluated the proposed method with six tree species. We carried out the evaluation experiment with various conditions, and found an optimal condition for our problem.
This research presents the character input system by using wearable camera based gaze estimation system (GES) for disorders. The GES is the system that can estimate the user's gaze direction. The proposed system uses the inside-out camera to get the user's gaze image. This system uses the gaze image to classify the character and create the word. Proposed system consists of two processes: Gazes estimation process and character recognition process. This research uses the Convolutional Neural Network (CNN) method to archive the good accuracy. By using the proposed method to train the model, the system was confirmed through evaluation experiments. The experimental results show the good accuracy of the proposed system compared with the previous research.