A dissertation discussion at Computer Science Department

A dissertation discussion at Computer Science Department (Localization of Strangeness from Crowded Video Scenes using 3D Convolutional Neural Network and Hybrid Descriptors Algorithm)

 It is proposed in this thesis presented by student Ali Abdul-Hussein Aziz. Two methods for hybridization After pre-processing work, and determining whether the video is crowded or not, the first method: mixed descriptor (local and global feature), to merging features, so that hybridization is not only to reduce the amount of data and time consumed but to create more understandable images and suitable for machine and human being perception.

The researcher reached several conclusions, including conducted experiments on two of data sets proved that the features learned in the proposed approach are very effective in the detection task, and a better execution time was obtained from the two models, and in terms of accuracy and loss, it is better than the 3D auto encoder unit. .

The discussion committee consisted of

• Dr. Mathil Imad Al-Din / the head.
• Dr. Hala Bahjat / Member / UOT.
• Prof. Dr. Rehab Falih / Member / UOT.
• Prof. Dr. Rahim Abdul-Sahib / Member / UOT
• Prof. Dr. Nadia Fadel / Member / Al-Nahrain University.
• Dr. Shayma Hamid / supervisor / UOT
• Dr. Iqbas Ei Al-Din / supervisor/ UOT.