%0 Generic %1 berman2019multigrain %A Berman, Maxim %A Jégou, Hervé %A Vedaldi, Andrea %A Kokkinos, Iasonas %A Douze, Matthijs %D 2019 %K %T MultiGrain: a unified image embedding for classes and instances %U http://arxiv.org/abs/1902.05509 %X MultiGrain is a network architecture producing compact vector representations that are suited both for image classification and particular object retrieval. It builds on a standard classification trunk. The top of the network produces an embedding containing coarse and fine-grained information, so that images can be recognized based on the object class, particular object, or if they are distorted copies. Our joint training is simple: we minimize a cross-entropy loss for classification and a ranking loss that determines if two images are identical up to data augmentation, with no need for additional labels. A key component of MultiGrain is a pooling layer that allow us to take advantage of high-resolution images with a network trained at a lower resolution. When fed to a linear classifier, the learned embeddings provide state-of-the-art classification accuracy. For instance, we obtain 79.3% top-1 accuracy with a ResNet-50 learned on Imagenet, which is a +1.7% absolute improvement over the AutoAugment method. When compared with the cosine similarity, the same embeddings perform on par with the state-of-the-art for image retrieval at moderate resolutions.

@misc{berman2019multigrain, abstract = {MultiGrain is a network architecture producing compact vector representations that are suited both for image classification and particular object retrieval. It builds on a standard classification trunk. The top of the network produces an embedding containing coarse and fine-grained information, so that images can be recognized based on the object class, particular object, or if they are distorted copies. Our joint training is simple: we minimize a cross-entropy loss for classification and a ranking loss that determines if two images are identical up to data augmentation, with no need for additional labels. A key component of MultiGrain is a pooling layer that allow us to take advantage of high-resolution images with a network trained at a lower resolution. When fed to a linear classifier, the learned embeddings provide state-of-the-art classification accuracy. For instance, we obtain 79.3% top-1 accuracy with a ResNet-50 learned on Imagenet, which is a +1.7% absolute improvement over the AutoAugment method. When compared with the cosine similarity, the same embeddings perform on par with the state-of-the-art for image retrieval at moderate resolutions.}, added-at = {2019-03-15T23:12:16.000+0100}, author = {Berman, Maxim and Jégou, Hervé and Vedaldi, Andrea and Kokkinos, Iasonas and Douze, Matthijs}, biburl = {https://www.bibsonomy.org/bibtex/29fb647b4f511705544e684f2f55d01d6/nmatsuk}, interhash = {75f93c73befe665aa8ab9b4488496bf9}, intrahash = {9fb647b4f511705544e684f2f55d01d6}, keywords = {}, note = {cite arxiv:1902.05509}, timestamp = {2019-03-15T23:12:16.000+0100}, title = {MultiGrain: a unified image embedding for classes and instances}, url = {http://arxiv.org/abs/1902.05509}, year = 2019 }