find-path

Find Path project finds humans paths and routes, such as sidewalks, park ways, forest paths. This project implements semantic segmentation approach. It uses VGG16 pretrained model.

Path Finding

Installation

TODO

Training

Go to calculations folder.

$ cd calculations

Run training.

$ python train.py

Run Trained Model

Check for examples in calculations/demo.py, calculations/demo.ipynb, calclulations/video_demo.ipynb files.

Dataset Maker

For making dataset, web based application was made which uses just JavaScript without any framework.

Installation

Before installation, make sure that NodeJS, npm and bower are installed.

$ cd dataset_maker
$ bower install

Launching

Open index.html and have fun.

Dataset

All dataset images have 320 width, 180 height and contain 3 channels. Every image has own .json file which describes object in the image. In this project only 3 classes are observed: boundaries (everything arround path), paths / ways and obstacles (things that are on path - eg. human, road pit and etc.). Dataset contains 300 images (I'll put a bit later).

Path Finding with OpenCV

Check calculations/cv/ folder.

Code References

https://github.com/MarvinTeichmann/tensorflow-fcn https://github.com/shelhamer/fcn.berkeleyvision.org https://github.com/machrisaa/tensorflow-vgg