Timezone: »

 
Poster
Recurrent Transformer Networks for Semantic Correspondence
Seungryong Kim · Stephen Lin · Sangryul Jeon · Dongbo Min · Kwanghoon Sohn

Wed Dec 05 07:45 AM -- 09:45 AM (PST) @ Room 517 AB #119
in Wed Poster Session A »

We present recurrent transformer networks (RTNs) for obtaining dense correspondences between semantically similar images. Our networks accomplish this through an iterative process of estimating spatial transformations between the input images and using these transformations to generate aligned convolutional activations. By directly estimating the transformations between an image pair, rather than employing spatial transformer networks to independently normalize each individual image, we show that greater accuracy can be achieved. This process is conducted in a recursive manner to refine both the transformation estimates and the feature representations. In addition, a technique is presented for weakly-supervised training of RTNs that is based on a proposed classification loss. With RTNs, state-of-the-art performance is attained on several benchmarks for semantic correspondence.

Keywords: Computer Vision · Visual Features
[ Paper] 

Author Information

Seungryong Kim (Yonsei University)
Stephen Lin (Microsoft Research)
Sangryul Jeon (Yonsei University)
Dongbo Min (Ewha Womans University)
Kwanghoon Sohn (Yonsei Univ.)

Related Events (a corresponding poster, oral, or spotlight)

More from the Same Authors