Switching to Discriminative Image Captioning by Relieving a Bottleneck of Reinforcement Learning.

Ukyo Honda, Taro Watanabe, Yuji Matsumoto

Browse the full WACV paper archive.