Optical flow and depth estimation
In this section, we will look at different ML tasks and the followed procedures to generate their corresponding ground truth.
Ground truth generation for computer vision
Computer vision aims at enabling computers to see using digital images. It is not surprising to know that vision is one of the most complex functionalities performed by our brain. Thus, imitating vision is not simple, and it is rather complex for state-of-the-art computer vision models.
Computer vision tasks include semantic segmentation, instance segmentation, optical flow estimation, depth estimation, normal map estimation, visual object tracking, and many more. Each task has its own unique way of generating the corresponding ground truth. Next, we will see samples of these tasks.
Image classification
The training images for this task usually contain one object, which is the object of interest. The annotation for this task is simply looking at each image and selecting...
