Photo Geolocation Estimation On Im2Gps

City level (25 km)

Continent level (2500 km)

Country level (750 km)

Median Error (km)

Reference images

Region level (200 km)

Street level (1 km)

Training images

Results

Performance results of various models on this benchmark

									Paper Title
ISNs (M, f*, S3)	43.0	80.2	66.7	-	0	51.9	16.9	4.7M	Geolocation Estimation of Photos using a Hierarchical Model and Scene Classification
PIGEOTTO	40.9	91.1	82.3	70.5	4.5M	63.3	14.8	4.5M	PIGEON: Predicting Image Geolocations
base (M, f*)	40.9	78.5	65.4	-	0	51.5	15.2	4.7M	Geolocation Estimation of Photos using a Hierarchical Model and Scene Classification
CPlaNet (1-5, PlaNet)	37.1	78.5	62.0	-	0	46.6	16.5	30.3M	CPlaNet: Enhancing Image Geolocalization by Combinatorial Partitioning of Maps
base (L, m)	35.0	79.7	64.1	-	0	49.8	13.5	4.7M	Geolocation Estimation of Photos using a Hierarchical Model and Scene Classification
Im2GPS ([L] KNN, sigma=4)	33.3	71.3	57.4	-	0	44.3	12.2	6M	Revisiting IM2GPS in the Deep Learning Era
Im2GPS (... 28m database)	33.3	73.4	61.6	-	28M	47.7	14.4	6M	Revisiting IM2GPS in the Deep Learning Era
StreetCLIP (Zero-Shot)	28.3	88.2	74.7	-	0	45.1	-	1.1M	Learning Generalized Zero-Shot Learners for Open-Domain Image Geolocalization
PlaNet (91M)	24.5	71.3	53.6	-	0	37.6	8.4	91M	PlaNet - Photo Geolocation with Convolutional Neural Networks
Im2GPS ([L] 7011C)	21.9	63.7	49.4	-	0	34.6	6.8	6M	Revisiting IM2GPS in the Deep Learning Era
PlaNet (6.2M)	18.1	65.8	45.6	-	0	30.0	6.3	6.2M	PlaNet - Photo Geolocation with Convolutional Neural Networks

0 of 11 row(s) selected.