CMSC733 Project 1: MyAutoPano

Dependencies

Python2.7 or Python3.5 -- Tensorflow 1.14.0 -- Numpy -- OpenCV -- ImgUtils

DataSet

You can download the MS-COCO dataset from here. Place your data on the Data folder alongside Code folder.

How to run

Phase 1

Unzip the file vasthana_p1.zip Mode into the folder unzipped.

Open ternimal and run the below commands

$ cd Phase1
$ cd Code
$ python Wrapper.py 

With the last commond, a argument has to be passed specifying the directory of the images.

$ python Wrapper.py --ImageDirectory="Mention path to your directory containing images"

Output image files

Program will generate various image output files in the CODE directory in Phase1 folder. If running various image data or test cases, it is recommended to take the back-up of generated output files before running the Wrapper.py again.

Phase 2

First go to Code directory and run gen.py to generate new data.

cd Code
python gen.py

After new data ha been generated, you can use following command to start training Supervised model.

python Train.py --CheckPointPath="../SupCheckpoints/" --ModelType="Sup" --MiniBatchSize=16 --LogsPath="SupLogs/"

Use following command to start training Unsupervised model.

python Train.py --CheckPointPath="../UnsupCheckpoints/" --ModelType="Unsup" --MiniBatchSize=16 --LogsPath="UnSupLogs/"

To run on test data, please make sure you have weights in the directory mentioned in the code. You can download weights from here

python Wrapper.py --Model="Unsup" --TestNumber=1 --MaxPerturb=32

For supervised model

python Wrapper.py --Model="Sup" --TestNumber=1 --MaxPerturb=32