Transparent Human Evaluation for Image Captioning
This work addresses the need for more transparent evaluation in image captioning, particularly for researchers and practitioners developing or assessing models, though it is incremental as it builds on existing evaluation practices.
The authors tackled the problem of evaluating image captioning models by introducing THumB, a rubric-based human evaluation protocol, and found that human-generated captions have substantially higher quality than machine-generated ones, especially in recall, while most automatic metrics incorrectly indicate the opposite.
We establish THumB, a rubric-based human evaluation protocol for image captioning models. Our scoring rubrics and their definitions are carefully developed based on machine- and human-generated captions on the MSCOCO dataset. Each caption is evaluated along two main dimensions in a tradeoff (precision and recall) as well as other aspects that measure the text quality (fluency, conciseness, and inclusive language). Our evaluations demonstrate several critical problems of the current evaluation practice. Human-generated captions show substantially higher quality than machine-generated ones, especially in coverage of salient information (i.e., recall), while most automatic metrics say the opposite. Our rubric-based results reveal that CLIPScore, a recent metric that uses image features, better correlates with human judgments than conventional text-only metrics because it is more sensitive to recall. We hope that this work will promote a more transparent evaluation protocol for image captioning and its automatic metrics.