Video-adverb retrieval with compositional adverb-action embeddings
About
Retrieving adverbs that describe an action in a video poses a crucial step towards fine-grained video understanding. We propose a framework for video-to-adverb retrieval (and vice versa) that aligns video embeddings with their matching compositional adverb-action text embedding in a joint embedding space. The compositional adverb-action text embedding is learned using a residual gating mechanism, along with a novel training objective consisting of triplet losses and a regression target. Our method achieves state-of-the-art performance on five recent benchmarks for video-adverb retrieval. Furthermore, we introduce dataset splits to benchmark video-adverb retrieval for unseen adverb-action compositions on subsets of the MSR-VTT Adverbs and ActivityNet Adverbs datasets. Our proposed framework outperforms all prior works for the generalisation task of retrieving adverbs from videos for unseen adverb-action compositions. Code and dataset splits are available at https://hummelth.github.io/ReGaDa/.
Related benchmarks
| Task | Dataset | Result | Rank | |
|---|---|---|---|---|
| Adverb-to-video retrieval | HowTo100M | mAP W56.7 | 7 | |
| Adverb-to-video retrieval | Adverbs in Recipes | mAP W70.4 | 7 | |
| Adverb-to-video retrieval | ActivityNet | mAP W23.9 | 7 | |
| Adverb-to-video retrieval | MSR-VTT | mAP (W)37.8 | 7 | |
| Adverb-to-video retrieval | VATEX | mAP W29 | 7 | |
| Video-to-Adverb Retrieval | HowTo100M | Acc-A81.7 | 7 | |
| Video-to-Adverb Retrieval | Adverbs in Recipes | Acc-A87.4 | 7 | |
| Video-to-Adverb Retrieval | ActivityNet | Acc-A77.1 | 7 | |
| Video-to-Adverb Retrieval | MSR-VTT | Acc-A (Adverb Retrieval)78.6 | 7 | |
| Video-to-Adverb Retrieval | VATEX | Acc-A0.817 | 7 |