An analysis into CAPTCHAs and how to break them using various machine learning algorithms. The effectiveness of naive Bayes, Probabilistic Principal Component Analysis (PPCA) and Feed Forward Neural Networks (FFNN) were compared. Given segmentation, a letter recognition accuracy of 84.11% was achieved using the FFNN.
My final report can be found here.
Conceptualized and constructed a remote control UAV. The UAV was modelled such that the lift, take off distance, and thrust would be able to carry a sufficiently high payload.
The constructed UAV was capable of carrying a payload of 8 tennis balls, 3 golf balls, and 8 ping pong balls. The total weight of the aircraft was 1.67kg.