Files in this item



application/pdfECE499-Sp2017-cheng-Yuan.pdf (1MB)
(no description provided)PDF


Title:Object Recognition on iCub Robot
Author(s):Cheng, Yuan
Contributor(s):Levinson, Stephen
Subject(s):object recognition
iCub robot
humanoid cognitive robot
Abstract:Object recognition is a technology for finding and identifying objects in an image or a video sequence. It relates to various fields including computer vision, machine learning, and image processing. Unlike other prior object recognition projects, this project mainly focuses on providing a visual recognition skill for iCub, a humanoid cognitive robot. A robot platform called YARP (Yet Another Robot Platform) works as a middleware to support the communication between hardware and software. This project uses supervised learning. Three different classes are trained and recognized by the robot: cup, sponge, and orange. For each class, 1800 sample images are collected from the cameras of the robot. Specifically, 80% of them are used for training and validation, and the remaining 20% are used for testing. The training model used for this project is neural network perceptron. Before feeding the features into the model, preprocessing steps are implemented on those features; these steps include clipping the image to smaller size (160*120 pixels), converting the image to greyscale, and applying a threshold. In the training process, randomized weights and constant bias are used as parameters. The Sigmoid function, the Relu function and the Softmax function, as well as backward propagation are used for training and updating class weights. The training results show that after 50 epochs the model converges to more than 90% accuracy. The testing results of the project show that the overall accuracy for the three classes is around 75% (results can be improved by follow-up modification). We can tell that the perceptron learning works well on this image set. Further modification can be done to increase the accuracy and the speed of training, such as using the stereo video to eliminate the noise and increasing the complexity of the neural network. This project can lead to more research such as feeding language signals to the robot or implementing reaching/picking after recognizing so that the robot will perform more like a human.
Issue Date:2017-05
Date Available in IDEALS:2017-08-18

This item appears in the following Collection(s)

Item Statistics