[3Rin4-75] Improvement of illustration image classfication using background removal
Keywords:Pseudo training data, Illustraion recognition, Image Recognition
In this research, we aim to develop a system for automatically classifying illustration images, using pseudo
training data. We used two types of pseudo-training data, one with the background of the photo data removed
and the other with the edge extracted after removing the background, and compare the differences in classification
accuracy. The model using the edge extracted data was more accurate than the model using the data with the
background removed. However, the model using the data with the background removed has lower accuracy than
the model using the photographic image as it is.
training data. We used two types of pseudo-training data, one with the background of the photo data removed
and the other with the edge extracted after removing the background, and compare the differences in classification
accuracy. The model using the edge extracted data was more accurate than the model using the data with the
background removed. However, the model using the data with the background removed has lower accuracy than
the model using the photographic image as it is.
Authentication for paper PDF access
A password is required to view paper PDFs. If you are a registered participant, please log on the site from Participant Log In.
You could view the PDF with entering the PDF viewing password bellow.