Skip to content

Detecting hand drawn flowcharts using Tensorflow Object Detection API - Faster RCNN

Notifications You must be signed in to change notification settings

Ruturaj123/Flowchart-Detection

Repository files navigation

Flowchart-Detection

Detecting hand drawn flowcharts using Tensorflow Object Detection API

This repository uses Tensorflow Object Detection API for detecting hand drawn flowcharts. The dataset images are located in the folder models/research/object_detection/images/test

I have used the Faster-RCNN Inceptionv2 model for detecting the flowcharts.

Output

Output

Screenshot_20190416_085315 Screenshot_20190416_085356

Running Locally

  • Install the Tensorflow Object Detection API

  • #From Flowchart-Detection/models
    export PYTHONPATH=$PYTHONPATH:`pwd`:`pwd`/research:`pwd`/research/slim:`pwd`/research/object_detection
  • #From Flowchart-Detection/
    python Object_detection_image.py        (#You can set any input image of your choice located in models/research/object_detection/images/test inside this script)
  • Coordinates of the bounding boxes will be stored in coordinates.json

NOTE: OpenCV is needed for displaying the image. It can be installed using pip install opencv-python