Caffe/OpenCV pipeline for classifying and describing scenes for cognitive research.
I made this as a work project to get some sensible images for a visual search-type experiment.
You can see how it works in the 'Scene_class.ipynb' notebook. All paths are absolute to my local machine at this point, and you will probably have to work to get all dependencies running - I used Ubuntu 16.04 as my OS.
- OpenCV3 (and dependencies)
- Python2.7 (I used Anaconda)
- Jupyter
- Caffe (and dependencies)
- pycaffe
These are downloaded from the Places2 database, I have used the test set, but you can use anything really.
- run: 'sh get_places2_test.sh' to get the ~20 Gb dataset.