Applied Deep Learning-Based Computer Visions in Agricultural Systems
POUL(AESC) 8270
3 hours
Applied Deep Learning-Based Computer Visions in Agricultural Systems
Course Description
Course will guide students to develop different deep learning algorithms for analyzing images or videos from agricultural systems. Students will be provided with methodologies for state-of-the-art deep learning algorithms for object detection and tracking, image classification, segmentation, and behavior recognition, meanwhile gaining hands-on programming skills via processing open-source datasets.
Athena Title
Deep Learning in Agriculture
Pre or Corequisite
[(STAT 6210 or STAT 6210E) and STAT 6220] or permission of department
Semester Course Offered
Offered fall
Grading System
A - F (Traditional)
Student learning Outcomes
Students will understand the techniques of start-of-art deep learning algorithms for images and video analysis in biological systems.
Students will gain hands-on Python programming experience in developing and applying the deep learning algorithms by using cloud computing or local high-performance computers.
Students will understand how to complete their own projects closely related to their own graduate research work using the learned techniques in this class.
Topical Outline
Concepts about convolutional neural network, deep learning, artificial intelligence, images, and videos
Deep learning-based image regression to predict continuous variables
Deep learning-based image classification to predict discrete variables
Deep learning-based object detection to localize objects of interests in images/videos
Semantic/instance segmentation to segment objects of interests in images/videos
Key point estimation to extract points and skeletons from target objects
Object tracking to depict trajectories and movements of individual objects
Behavior recognition to classify behaviors of animals or other agriculture species from videos
Other useful deep learning algorithms for biological systems, including image searching, image generation, hyper resolution, and image captioning
