果冻影院

A robot鈥檚 environment is often composed of objects that need to be interacted with or avoided. The problem of estimating the geometric properties of these objects (size and shape) were considered in the core module 鈥淐omputer Vision and Sensing鈥�. However, in many cases semantic information 鈥� such as the type of object or the capabilities and properties of it 鈥� need to be determined as well.

This module will look at how to develop algorithms to predict semantic properties from raw sensor data and computed 3D models. The module builds on the Term 1 module, Machine Learning for Robotics because this is the most dominant and successful approach to this problem area.

Aims:

The aims of the module are to:

• Provide students with an understanding of the challenges of identifying and classifying objects.
• Provide students with an understanding of current algorithms and their strengths and limitations.
• Support students to acquire experience in implementing and evaluating these algorithms.

Intended learning outcomes:

On successful completion of the module, a student will be able to:

1. Identify and recognise the main concepts related with the challenges of object detection and classification.听
2. Recognize and evaluate the strengths and limitations of different algorithms in object identification and classification.
3. Apply appropriate evaluation techniques to assess the performance of implemented algorithms.
4. Analyze and interpret the results obtained from evaluating algorithms.
5. Apply responsible practices in handling data, ensuring privacy, and avoiding biases.
6. Critically assess the performance and limitations of algorithms in different scenarios.

Indicative content:

The following are indicative of the topics the module will typically cover:

• Image classification (CFAR10 / NIST.)
• Semantic segmentation.
• Object detection.
• Object classification. (YOLO, R-CNN.)
• Performance evaluation.
• Depth from a single image.
• Other sensing modalities (LiDAR, etc.)
• Out of distribution samples and errors.

Requisite conditions:

To be eligible to select this module as optional or elective, a student must be registered on a programme and year of study for which it is formally available.