This visual classification model classifies photos to one of 18 visual attributes which are intended for the measurement of touristic destination image.
This conference paper introduces the model.
It is fine tuned on touristic destination photography from the BEiT-L model trained on ImageNet21k.
For validation, we evaluated with a ground truth dataset of 1800 photos (100 per visual attributes) and achieved 95% accuracy. The ground truth dataset is publicly available for benchmarking other models against ours.
Model weights are made available under the Creative Commons Attribution Non Commercial Share Alike 4.0 license.
- Downloads last month
- 0