Progress Report on the Development of an autonomous F450 quadcopter-based sentinel capable of automatic landing and recharging (Week 7) Name: Paul Okewunmi Date: 7th January 2024 Paul Okewunmi Work Plan Workplan: I. Ongoing tasks ● Development and testing of machine learning models to automatically detect events and objects of interest II. Report on activities during the week ● I completed and tested the pipeline I came up with for the vehicle colour classification, In summary, it begins with detecting vehicles and identifying their bounding boxes, then crops the vehicle out of the frame, extracts its dominant colour using Kmeans clustering, processes HSV values, and finally predicts the vehicle's colour class. Output Video: This video shows the result of the colour classification, It can be observed that the model struggles with the colour brown, and instead misclassifies it as blue, this can be attributed to class imbalance, the dataset I gathered contained fewer samples for the brown class, and this can be solved by gathering more samples. III. Self-Appraisal Last week’s indicators were partially met due to the progress in classification with vehicle detection, but some work still needs to be done regarding the video transmission to enable this to translate well to SN2 (sn2 video example) IV. Proposed next steps ● Improving colour classification accuracy and more data gathering. ● Following our midweek conversation, there’s a need for a dataset of vehicles and their colours from a top-down drone view. So plans will be put in place to create one and I’ll be submitting a first draft of the execution of this task on Monday. V. Indicators. ● A complete work plan for the data gathering task, detailing needed resources and objectives.