mrcnnModuleService

A module build on the matterport/Mask_RCNN Repository providing the use of the neural network module in a separate Process for easy Multiprocessing Implementation.

Imports required:

Mostly I used the same build as like the MRCNN Repository and the Tutorial on Training your own neural network(see below).

• MRCNN Repository: https://github.com/matterport/Mask_RCNN (Download the mrcnn directory and copy it in your Project directory)

other Dependencies:

• tensorflow 1.14
• keras 2.2.4
• numpy
• scipy
• Pillow
• cython
• matplotlib
• scikit-image
• opencv-python
• h5py
• imgaug
• IPython[all]

how to use:

• Make sure you have the Imports/Dependencies above
• Copy the python file mrcnnModuleService in the same directory as like the mrcnn directory
• Import in your Code: from mrcnnModuleService import NeuralNetService (if you do not want a separate Process, import the NeuralNet class)
• Create a new network Configuration(example found in mrcnnModuleService I used in a different Project), should be similar to the Configuration your network was trained with
• Create a new NeuralNetService-Object, as params you give the Configuration and the absolute Path to the loading weights
• Start the Process using the start_service() Method
• Add a prediction task with add_prediction_task(imgList)
• Get the corresponding Predictions with get_preditcion_task()
• Shut down the Service if you do not need it anymore with shut_down_service()
• To restart the Service you need to initiate it first: init_service(config, path) and start_service()
• look at the Description in the Methods for more info

other interesting Stuff:

• Tutorial on Training your own Mask_RCNN Network: https://github.com/akTwelve/cocosynth/blob/master/notebooks/train_mask_rcnn.ipynb