Data Lens

Visualise your dataset before training the model in one line!

Made changes to the bounding boxes or images? Save time by visualising the data and avoid mistakes before starting the training process

Use Case:

• Sanity check for the right dataset and annotations
• Dont forget to resize the bounding boxes
• Visualise what augmentation functions do to your data

Using simple GUI, visualise random images for your tasks

Installation (Python 3.6+):

pip install datalens

Usage:

import datalens
datalens.Visualise(image_dir_path = image_dir_path, annotations_dict = annotations_dict, count = count)

Current version supports Object Detection in 2D images.

Data Formatting:

image_dir_path = #PATH TO IMAGE DIRECTORY
annotations_dict = #OBJECT DETECTION - {image_filename: [{"bbox": [x, y, width, height], "category_id": <int>}, ...], ...}
count = 10 #NUMBER OF IMAGES TO VISUALISE

Contributions are welcome for different machine learning tasks for text, images and 3D Point cloud data.