What five categories can computer vision be divided into?

Computer vision is a research field covering many technologies and applications. The following are five common computer vision tasks:

Image classification:

Image classification refers to the classification of images into different categories according to their contents. This is one of the most basic tasks in computer vision, which involves feature extraction and pattern recognition. Deep learning technology, especially convolutional neural network (CNN), has made remarkable achievements in this field.

Object detection (object detection):

Target detection not only needs to identify the target category in the image, but also needs to determine the position and bounding box of the target. This task usually involves dealing with target location and classification at the same time. Common target detection methods include R-CNN, YOLO, SSD, etc.

Semantic segmentation:

Semantic segmentation is to assign every pixel in the image to a corresponding category, so as to accurately divide different objects in the image. This topic is widely used in automatic driving, medical image analysis and other fields. Common semantic segmentation methods include FCN, U-Net, DeepLab and so on.

Instance segmentation:

On the basis of semantic segmentation, case segmentation further distinguishes different cases of the same category. This is very important for understanding the number and relationship of objects in the scene. Common examples of segmentation methods are Mask R-CNN, SOLO and so on.

Posture estimation:

Attitude estimation refers to estimating the spatial attitude of objects from images, such as the detection of key points of human body, the estimation of object position and attitude, etc. This kind of task is widely used in motion recognition, augmented reality, robot navigation and other fields. Common attitude estimation methods include OpenPose, AlphaPose, POSEC3D and so on.

These five types of computer vision tasks cover many specific application scenarios and promote the development and innovation in the field of computer vision.