Speaking the Same Language: Matching Machine to Human Captions by Adversarial Training
This addresses the problem of generating more human-like and diverse captions for image captioning systems, which is incremental as it builds on existing adversarial training methods.
The paper tackled the problem of machine-generated image captions being distinct from human captions due to deficiencies in word distribution, vocabulary, and bias, by changing the training objective to generate captions indistinguishable from human ones using adversarial training and an approximate Gumbel sampler. The result was a set of diverse captions that achieved comparable correctness to state-of-the-art methods while being significantly less biased and better matching word statistics.
While strong progress has been made in image captioning over the last years, machine and human captions are still quite distinct. A closer look reveals that this is due to the deficiencies in the generated word distribution, vocabulary size, and strong bias in the generators towards frequent captions. Furthermore, humans -- rightfully so -- generate multiple, diverse captions, due to the inherent ambiguity in the captioning task which is not considered in today's systems. To address these challenges, we change the training objective of the caption generator from reproducing groundtruth captions to generating a set of captions that is indistinguishable from human generated captions. Instead of handcrafting such a learning target, we employ adversarial training in combination with an approximate Gumbel sampler to implicitly match the generated distribution to the human one. While our method achieves comparable performance to the state-of-the-art in terms of the correctness of the captions, we generate a set of diverse captions, that are significantly less biased and match the word statistics better in several aspects.