Skip to content

Latest commit

 

History

History
15 lines (10 loc) · 1.42 KB

README.md

File metadata and controls

15 lines (10 loc) · 1.42 KB
Raw

Semantic Preditors

All customized semantic predictors can be placed under this folder, derived from base_predictor.py and registered in predictor_settings.py. In this repo, scene-related predictor is provided.

Scene Attribute and Category Prediction Model

This model is borrowed from places365 repository.

Before using this predictor, please download following files:

  • Definition of 365 categories on Places365 dataset. Please save file categories_places365.txt under annotations folder.
  • Definition of 102 attributes on Places365 dataset. Please save file labels_sunattribute.txt under annotations folder.
  • Pre-trained weight for category prediction. Please save file wideresnet18_places365.pth.tar under pretrain folder.
  • Additional weight for attribute prediction. Please save file W_sceneattribute_wideresnet18.npy under pretrain folder.
  • Existing file scene_wideresnet.py in this folder is borrowed from here, yet slightly changed (Line 152 and Line 156).