Single image method finds repeated objects without prior training data
Bottom-up Modeling of Repeated Elements via Single Image Analysis-by-Synthesis
Computer Vision and Pattern Recognition
Summary
Finding repeated objects or patterns in images usually needs many labeled examples or special image collections. This paper shows it is possible to teach a computer to recognize and copy repeated parts using only one single image. The method works by creating a simple 2D model of these repeated objects from the image itself, without needing any pre-existing knowledge. Tests on a variety of real images show it can find consistent repeated elements and handle differences among them better than other common approaches. This suggests that understanding repeated elements in images can be done from a single picture alone.
object discoverysingle image analysisimage reconstructionprototype learningbottom-up modelingintra-category variation2D image synthesisunsupervised learningimage decompositionFSC-147 dataset
Authors
Syrine Kalleli, Alexei A. Efros, Mathieu Aubry
Abstract
We address the problem of discovering repeated elements from a single image. In contrast to existing approaches that depend on large annotated datasets, curated multi-image collections, or object segmentation masks, we show that a single image can suffice to learn a meaningful object model in a completely bottom-up fashion, without any prior knowledge beyond a coarse scale prior. Our method learns a tunable image-space prototype of the repeated elements through a reconstruction objective, enabling the model to identify and synthesize consistent object instances within the same image. Experiments on 116 real images from the FSC-147 dataset demonstrate that our method successfully learns coherent element models and captures intra-category variation on challenging images. Qualitative results reveal superior reconstructions and interpretable decompositions compared to classical decomposition, joint alignment, and 3D object modeling methods, while maintaining a simple 2D formulation. These results suggest that meaningful object discovery can emerge from single image learning alone.