Faculty of ICT Department of Intelligent Computer Systems CSA3803 (Advanced Project) Proposal Form Title: Object localization and recognition Project Supervisor: Dr George Azzopardi Project Co-supervisor: (if applicable) Main Subject Area: / Brain-inspired computer vision Object recognition is an important topic in the field of computer vision. Recently, a novel method called COSFIRE has been found very effective in various applications including, traffic sign recognition, bifurcation detection in and retinal images, among others. This approach is based on time consuming orientation-selective (Gabor) filters. In this project, we will investigate the using SIFT keypoint descriptors as input to the COSFIRE approach. It is expected to Detailed Description of Task (c. achieve better effectiveness and better efficiency in the localization and 200 words): recognition of patterns of interest in images. We will evaluate the method on the benchmark data set Flickr logos which can be accessed from http://www.multimedia-computing.de/flickrlogos/ The Matlab implementation of the COSFIRE approach can be downloaded from Matlab script: http://tinyurl.com/p5gklaw Resources Available: Computer G. Azzopardi and N. Petkov, “Trainable COSFIRE filters for keypoint detection Recommended Reading (at and pattern recognition”, IEEE Transactions on Pattern Analysis and Machine least one title): Intelligence, vol. 35 (2), pp. 490-503, 2013. • Prerequisite Knowledge Required: Download: http://tinyurl.com/mdb6gmk Interest in computer vision Indication of any Ethical Issues and How these will be Tackled N/A (if applicable): _____________________________ ________________________ Signature (Supervisor) Date _____________________________ ________________________ Signature (Student) Date Faculty of ICT Department of Intelligent Computer Systems CSA3803 (Advanced Project) Proposal Form Title: TensorFlow for websites Project Supervisor: Prof. Alexiei Dingli Project Co-supervisor: (if applicable) Main Subject Area: The student will make use of a new Machine Learning library, which has just been released by Google. TensorFlow is a powerful library for doing large-scale numerical computation. One of the tasks at which it excels is implementing and training deep neural networks, a new breed of machine learning techniques. It proved to be extremely successful on the MNIST database which consists of images of handwritten digits like these: Detailed Description of Task (c. It also includes labels for each image, telling us which digit it is. For example, 200 words): the labels for the above images are 5, 0, 4, and 1. For this project, you are requested to perform a similar task but with websites. So you will: 1. Make use of TensorFlow (as per tutorial) to create a system to categorize images. 2. Make use of Phantom JS to take snapshots of websites. 3. Make use of website rankings such as (https://www.woorank.com or http://www.alexa.com/topsites) to get online rankings. 4. Train the classifier created in 1 with the websites Resources Available: Recommended Reading (at least one title): http://tensorflow.org/ http://tensorflow.org/tutorials http://tensorflow.org/get_started/os_setup.md http://phantomjs.org/screen-capture.html Prerequisite Knowledge Required: Machine Learning Indication of any Ethical Issues and How these will be Tackled No ethical issues (if applicable): _____________________________ ________________________ Signature (Supervisor) Date _____________________________ ________________________ Signature (Student) Date