IcoCap: Improving Video Captioning by Compounding Images

Benchmark Model Rank Results
video-captioning-on-msr-vttIcoCap (ViT-B/16)–CIDEr: 60.2METEOR: 31.1ROUGE-L: 64.9BLEU-4: 47.0
video-captioning-on-msr-vttIcoCap (ViT-B/32)–CIDEr: 59.1METEOR: 30.3ROUGE-L: 64.3BLEU-4: 46.1
video-captioning-on-msvdIcoCap (ViT-B/16)–CIDEr: 110.3BLEU-4: 59.1METEOR: 39.5ROUGE-L: 76.5
video-captioning-on-msvdIcoCap (ViT-B/32)–CIDEr: 103.8BLEU-4: 56.3METEOR: 38.9ROUGE-L: 75.0
video-captioning-on-vatexIcoCap (ViT-B/16)–BLEU-4: 37.4CIDEr: 67.8METEOR: 25.7ROUGE-L: 53.1
video-captioning-on-vatexIcoCap (ViT-B/32)–BLEU-4: 36.9CIDEr: 63.4METEOR: 24.6ROUGE-L: 52.5