Data Collection and basic Filtering

For small tasks or in initial phase of a project we can go about curating data using Google. If we are downloading dataset from google it needs some basic filtering in place.

Code above shows how we can use simple ssdlite mobilenet to filter and crop images for furnitures.

Whole process can be described in four steps

Step 1: Create your list of objects

Step 2: Data Collection: Download the images from Google

Step 3: Load the graph and weight file

Step 4: Cleaning: Crop and filter the downloaded image

Start a tensorflow session and apply the classifier on each downloaded image file. Get all the detected objects along with bounding boxes.

For each detected object check if the classId is same as the object in consideration. For example if the downloaded image are of chair then detected objects in the image should have classID 62. Crop all such objects from the image and save them. Delete original image.

Full code below:

Related posts:

Joe Biden is sloppy. He makes comments without fully considering the context or how they might be interpreted by others. Biden speaks off the cuff more often than a politician ever should. A comment…

Using data analysis and machine learning to find the ideal location for any given restaurant in Toronto

How to choose a Career? A career planning is a strategy that you continuously develop to manage your learning and progression. There are so many options that help you to choose a career for your…