SPIKE dataset Files

**UPDATE: As CNTK becomes older, more packages, libraries and versions of python etc. seem to be incompatible with each other. Python 3.7 is not usable for this tool and Python 3.6 may not be either.**


This folder contains the two main components necessary for detecting spikes in images with our trained model. 

These components are; model files, containing our trained networks and .py files, containing the settings and parameters we have used with the Faster R-CNN implementation from the freely available Microsoft Cognitive Toolkit (CNTK), found at the following link https://www.microsoft.com/en-us/cognitive-toolkit/

Some things to remember / steps to follow are:

- New images for testing must be placed in a folder named 'positive' in .../CNTK/Examples/Images/DataSets/

- This folder also requires five text files with information about the dataset, these text files can be generated running the following file with cmd, .../CNTK/Examples/Image/Detection/utils/annotations/annotation_helper.py . Toward the end of the annotation_helper.py file a line of code needs to changed to represent your testing data: data_set_path = os.path.join(abs_path, "../../../DataSets/YOUR_DATA")

- The pre-trained model needs to be placed in the folder  ...CNTK/Examples/Images/Detection/FasterRCNN/Output

- The necessary VGG16 model can be downloading by executing the download_model.py script in the ...CNTK/PretrainedModels/ folder. To download VGG16 only (recommended) use the following command: python download_model.py VGG16_ImageNet_Caffe

- Replace the files in ...CNTK/Examples/Image/Detection/FasterRCNN/ with those provided above. Do the same  with the files in ...CNTK/Examples/Image/Detection/utils/configs/

- Finally execute the file: python run_faster_rcnn.py

Some other things to note:

- As CNTK becomes older more packages, libraries and versions of python etc. seem to be incompatible with each other. Python 3.7 is not usable for this tool and Python 3.6 may not be either.

- Using pip will make installation of libraries and general running of the tool much easier

- CNTK requires various modules such as numpy, scipy, easydict etc. that will need to be installed as you progress through the steps of running the tool. This is easy to do with pip e.g. pip install numpy. You will be prompted each time a new module is required