You will be surprised to know the release of two new public data set which is subdomain off the open image dataset. Two subdomains are integrating with open image data set name shellfish object detection and vehicles object detection. From the vast landscape of the open image data set, shellfish and vehicles are just small windows that are meant to provide minor evidence of data sets that can be easily constructed with open images.
Object detection is one of the most exciting aspects which you can find in computer vision. It allows the user to detect different objects of an individual like an image, locate its size and position. Nowadays, most object detection models are powered by a deep learning network in the training data set.
You may find different formats for preparing and annotating data set for object detection. but the most popular formats are
- Pascal VOC
- Microsoft COCO
You can also choose to create your own data set. Let’s assume when an autonomous vehicle collects the data. In the very first step, you have to fix the camera to the vehicle so that you can record the video while the vehicle is moving. From this; you can get a video file which is a combination of various frames. You can divide the video file into frames by writing a few lines of code in Python.
Introduction to data annotation tool
Under this, we start labeling the objects in the image. image annotator is the best tool that can be used to draw the bounding boxes in the image. It also allows you to add textual information for the objects in the image. By using these label data, you can train your deep learning model.
This is how you can create custom open image data set for object detection.