 |
Lecture on Deep learning models for computer vision An overview of convolutional neural networks and their applications, from image classification, semantic segmentation, object detection, instance segmentation, visual object tracking, optical flow estimation, to video classification. Details on architectural design, novel concepts, advantages and disadvantages. Conclusion with the recent vision transformers. ECS709 - Introduction to Computer Vision (Winter 2022) at QMUL Lecturer: Prof. A. Cavallaro |
 |
Lecture on Interest Points Following up the basics of detection and description of interest points, the lecture goes in depth of local image features and covers their applications and taxonomy, the more advanced descriptors related to the SIFT family (RootSIFT, DSP-SFIT), binary descriptors (BRIEF, BRISK, ORB, LATCH, MORB, BOLD) with their principle designs (pairwise and triplet comparisons, sampling pattern design, orientation assignment methods, instability of binary tests, multi-scale), deep learning patch-based descriptors (DeepCompare, DeepDesc, TFeat, L2-Net, HardNet, DeepBit) with details on the design losses (contrastive, triplet, mining hard negatives, anchor swap, progressive sampling, unsupervised learning) and architectures, and deep learning image-based approaches (LIFT, LF-Net, SuperPoint, D2-Net, R2D2). ECS709 - Introduction to Computer Vision (Winter 2022) at QMUL Lecturer: Prof. A. Cavallaro |
 |
Tutorial on How to review scientific papers A 2-session tutorial, titled "The reviewer doesn't understand & is wrong! Transitioning into the reviewer's shoes", for students that just started their PhD or wanted to gain more insights on the reviewing process and guidelines towards the preparation of high-quality reviews. Aims of the tutorial include understanding the reviewer role and reviewing process as an intermediate step for better writing research papers, mindset shift to be the reviewer of your own papers and colleague papers, and how to structure constructive and polite feedback for high-quality reviews with two 4-page published papers. The tutorial includes tips and guidelines from existing resources in journals, conferences, and blogs. The tutorial was delivered internally to the research group, and was followed by a throughout and individual feedback on the written reviews. |
Demonstrator (former Postgraduate Teaching Assistant) at QMUL
ECS709 - Introduction to Computer Vision (2018/2019, 2019/2020)
Lecturer: Prof. A. Cavallaro
Number of students: 66 (2019/2020)
Number of students: 33 (2018/2019)