A Tensorflow Implementation of Mask-RCNN

This repository contains an implementation of Mask-RCNN based on the Matterport implementation. The network is fully developed in tensorflow 1.14 with no keras function inside.

Prerequisites

The required libraries are:

• python = 3.6
• tensorflow-gpu = 1.14

And the following with the latest compatible version:

• numpy
• scipy
• matplotlib
• pillow
• pycocotools
• cython
• IPython[All]
• Imgaug

Running the tests

Run on Jupyter the Demo.ipynb file to check how to detect and test the time performances.

Keras Version Compatibility (convert_weights.py)

In order to allow a soft passage from Keras version to Tensorflow version, we develop a script to convert your .h5 weights file in tensorflow checkpoints. The generated checkpoints will perfectly reproduce the performances of your keras version model.

Available dataset and prepare data scripts

This repository has been developed in the context of Second Hands European project and an associated dataset had been created: https://github.com/alcor-lab/SecondHandsDataset This repository also offers a sample script to prepare data for training.

Authors

• Edoardo Alati (Edwyn00) & Malik Bekmurat (bekmake)

Acknowledgments

• Matterport Company for their first Keras implementation at https://github.com/matterport/Mask_RCNN