20261_1

Inferring where an image was taken from its visual content alone is a challenging problem at the intersection of computer vision and geospatial reasoning, owing to the visual similarity of distant places and the vast space of possible locations. A single street-level image contains numerous geographic cues: the language on signs, license plates, the side of the road traffic drives on, vegetation, the position of the sun, and architectural style. Combining these cues can lead to a good estimation of the Location.

This work investigates a multimodal approach that makes these cues explicit. A ResNet backbone, pretrained on ImageNet and fine-tuned on a purpose-built dataset of geotagged street-level images, learns location-discriminative visual representations. A second model is trained to combine the result of the first model together with complementary signals, like recognised text or Architecture. The output is a final location estimate.

(Felix Kreuer und Kaan Sarac / Prof. Fu, SoSe 2026, bestes Poster)