Visual Data Extractor using TensorFlow

The project was tested successfully on Python 3.7.9 and TensorFlow 2.3.1 respectively.

It is assumed that you already have following things installed on your machine.

• Python 3.x
• TensorFlow 2.x
• OpenCV
• NumPy

If not already, kindly install these first before you move ahead.

Follow the steps below to train an end-to-end custom object detection model for detecting visual data in images/dashboards.

Steps:

1. Clone the repository on your local machine.

2. Upload the Visual_Data_Extractor_using_TensorFlow.ipynb notebook file on Google Colab and run the cells one-by-one by following the instructions.

3. Install the TensorFlow Object Detection API (if not already) using pip install tensorflow-object-detection-api.

4. Open the label_map_util.py file and Edit Line 132 by replacing with tf.gfile.GFile(path, 'r') as fid: with with tf.io.gfile.GFile(path, 'r') as fid:. You will find this file in the following path: C:\Users\<your-username>\AppData\Local\Programs\Python\Python37\Lib\site-packages\object_detection\utils

5. Once the model is trained completely, download and extract the saved_model.zip file inside the Visual-Data-Extractor-using-TensorFlow folder. You will already find a pre-trained model here.

6. Open command prompt and run py detect_graphs.py file.

Facts:

• The model was trained for 5 classes of visualizations viz. Pie Chart, Bar Graph, Donut, Line Chart and Area Chart respectively.
• Take a look at the dataset (you'll need to extract the .zip file) used for training this model to get an idea about the directory structures.

Output: