Make python environment with:
* python3 -m venv myenv
* source myenv/bin/activate
* pip install -r requirements.txt




==============================================
Appendix A Experiment
==============================================
Same applies to FashionMNIST and CIFAR-10
* Model weights included
* Run 'Data Collection.py' to get data for comparisons
* Run 'Expt1.py' to save the figs and generate the data




==============================================
Experiment 1/2:
==============================================
Download the relevant datasets into a 'data' folder:
CUB-200: http://www.vision.caltech.edu/visipedia-data/CUB-200-2011/CUB_200_2011.tgz          ---- Use standard training and testing split
ImageNet: https://www.kaggle.com/c/imagenet-object-localization-challenge                ---- Use Training and Validation Data

for CUB-200 have your data directory like:
	data/
		CUB_200_2011/
			images/
			etc...

Run the scripts



==============================================
Experiment 3:
==============================================
* Use whatever survey making interface you like.
* Sample materials from 'Sample Materials/sample materials.ipynb'. Refer to the main paper for the study layout.
* You have the option to reuse the materials here, or sample your own.
* Crowdsource the data with two counterbalanced groups (N=163, 81/82) as described in the main paper.




==============================================
Section 2 "Speeding Up Twin-Systems" Experiment:
==============================================
* Setup ImageNet dataset in 'data' folder in directory as before in Experiment 2
* Add contributions derived in Experiment 2
* Run main.py




