[code & report] This is an Implementation of Image SuperResolution using standard CNNs as well as CNNs containing Residual Layers. To compare the different interpolation techniques used while upsampling our images, the SRCNN has been used to make direct comparisons between Bicubic, Bilinear and Nearest Neighbour interpolations
[code & report] An implementation of a Food Calorie estimator using YOLOv5 on the ECUSTFD food dataset. This model takes into account bounding box width and height taken from both top and side views as well as the food item class using YOLOv5 for object detection. Using these as input parameters, we train a simple ANN to give the calories contained as a regression output.
[code & report] Implemented a Bayesian Neural Network to classify a simple noisy XOR dataset. The model can be extended to work with more complex datasets. Defined posterior and likelihood functions and used Markov Chain Monte Carlo sampling, using the Metropolis‑Hastings algorithm.