A Framework for an automatic evaluation of image description based on an image accessibility guideline
MetadataShow full item record
In this digital world, it is mandatory to have equal access for all the people to understand the images. Inaccessible image descriptions can create barriers for some users to access the images. This paper presents a framework to automatically evaluate image descriptions under the consideration of NCAM image accessibility guidelines. Different research papers of image descriptions and image accessibility were studied for the literature review to explore more about the methodology to accomplish the evaluation of image descriptions. Machine learning is used to build a model which predicts how accessible an image description is with respect to NCAM guidelines. Random forest model was trained using Flickr8K dataset, and the dataset was labeled according to the NCAM guidelines. Standard error and Accuracy were used as a metric to calculate the accuracy and performance. Both the quantitative and qualitative evaluations found that the proposed framework is as effective as human evaluation. The framework is believed to be helpful to the web authors to check the accuracy of any image description with NCAM guidelines in order to make accessible images for the web content.