This repo contains VPR models that have been fine-tuned for indoor usage. Currently available models come from the following repositories:

• EigenPlaces: Training Viewpoint Robust Models for Visual Place Recognition;
• Optimal Transport Aggregation for Visual Place Recognition (Salad).

The fine-tuned procedure is inspired by Deep Visual Geo-localization Benchmark

The indoor datasets used for fine-tuning are:

• Gangnam Station B1-B2 and Hyundai Department Store B1-1F-4F;
• Baidu Mall.