machine learning - Training on MNIST, detecting sequences of numbers? -


after you've trained model on mnist set, how can classify image having 2 digits? more generally, how train model detect number of digits on image?

there hot field called "object detection" tries want. in general, can detect (digits, people, cars, etc) images , videos.

the state-of-the-art techniques fall 2 categories:

  1. faster-rcnn, first proposes lot of candidate windows objects of interest , detects inside these windows.
  2. ssd, scans images once , detect objects, faster not reliable compared faster-rcnn.

a well-known real-time object detection method yolo (you once), falls in ssd category, , has impressive real-time demo here, give sense of object detection. search these methods' names , find lot of example code satisfies needs.

if looking digit detection, check out work surrounding stanford's house number dataset. however, note these works 5 , more years ago , not beat general methods faster-rcnn , ssd.


Comments

Popular posts from this blog

ios - MKAnnotationView layer is not of expected type: MKLayer -

ZeroMQ on Windows, with Qt Creator -

unity3d - Unity SceneManager.LoadScene quits application -