VSE++: Improving Visual-Semantic Embeddings with Hard Negatives

18 Jul 2017  ·  Fartash Faghri, David J. Fleet, Jamie Ryan Kiros, Sanja Fidler ·

We present a new technique for learning visual-semantic embeddings for cross-modal retrieval. Inspired by hard negative mining, the use of hard negatives in structured prediction, and ranking loss functions, we introduce a simple change to common loss functions used for multi-modal embeddings. That, combined with fine-tuning and use of augmented data, yields significant gains in retrieval performance. We showcase our approach, VSE++, on MS-COCO and Flickr30K datasets, using ablation studies and comparisons with existing methods. On MS-COCO our approach outperforms state-of-the-art methods by 8.8% in caption retrieval and 11.3% in image retrieval (at R@1).

PDF Abstract

Datasets


Results from the Paper


Task Dataset Model Metric Name Metric Value Global Rank Benchmark
Cross-Modal Retrieval Flickr30k VSE++ (ResNet) Image-to-text R@1 52.9 # 12
Image-to-text R@10 87.2 # 12
Image-to-text R@5 80.5 # 12
Text-to-image R@1 39.6 # 11
Text-to-image R@10 79.5 # 12
Text-to-image R@5 70.1 # 11

Methods


No methods listed for this paper. Add relevant methods here