January 2018
Beginner to intermediate
284 pages
8h 35m
English
Evaluation of the image caption/description results is not an easy task. Similar to machine translation, some known challenges include, but are not limited to:
Often, two major aspects are considered—linguistic quality and semantic correctness, and two types of approaches are commonly used—human judgments and automatic measures.
The human evaluation is often used through crowdsourcing, for example Amazon Mechanical Turk. The tasks are often designed to include questions on grammar, content, and fluency. Judgers are provided with images together with the generated descriptions. In some other cases, ...
Read now
Unlock full access