Weekly project deliveries for the Module C6 based on Video Analysis, from Master's in Computer Vision coordinated by Universitat Autònoma de Barcelona (UAB).
-
Week 1: the goal is to explore several approaches for background detection, using 2024 NVIDIA AI City Challenge database. The goal is create a model to differentiate between the static background and moving cars.
-
Week 2: The goal of this week is to finetune a detection model and perform tracking of the cars of the video sequence s03_c010 from the 2024 NVIDIA AI City Challenge database and evaluate the results using the metrics HOTA and IDF1 from TrackEval.
-
Week 3: The goal of this week is to implement optical flow (OF) to the previous weeks detection and tracking algorithm trying to improve the results obtained.
-
Week 4: The goal of week 4 is to perform a multicamera car tracking algorithm using the techniques learned in any of the previous weeks using any of the video sequences from the 2024 NVIDIA AI City Challenge database.
Final Presentation: https://docs.google.com/presentation/d/1xBH7wxrVYxIpcjhYRKThHyvfbkOVPQRe1fW0kLkclL8/edit?usp=sharing
-
Week 5: The goal of this week is to get familiar with the baseline and dataset to perform action classification in football clips.
-
Week 6: The goal of this week is to precisely spot the football (ball) actions in time (time instants) without altering the temporal resolution.
-
Week 7: The goal of this week is to precisely spot the football (ball) actions in time (time instants) with no constraints.
Final Presentation: https://docs.google.com/presentation/d/1Jn2C2AdgFuSywCQU5jujKKjlYpUJ2CwLcqrxL1GCaxM/edit?usp=sharing