Ian Padovani's repositories
DNN-Speech-Recognizer-Udacity
Udacity project. Trains multiple architectures for ASR acoustic models and compares them.
facial-keypoint-detection-udacity
Udacity project. Uses a CNN to detect facial keypoints in an image. Combines this with Haar Cascades to isolate faces in an image and then detect their keypoints.
HMM-POS-Tagger-udacity
Udacity project. Compares two POS tagging models: A most-frequent-class tagger and a Hidden Markov Model.
machine-translation-udacity
Udacity project. Machine translation from English to French on a toy dataset using a DNN with embedding and bidirectional RNN layers.
dog-breed-classification-udacity
Udacity project. Uses transfer learning to create a dog breed classifier from VGG16. Takes in an image and returns the dog breed or what dog a person looks like.
face-generation-udacity
Udacity project. Trains a Generative Adversarial Network (GAN) to generate faces, based off a subsection of the Celeb-A dataset.
image-captioning-udacity
Udacity project. Uses a CNN as a feature extractor to analyse images and an RNN for generating captions to those images. Results are fair but the model seems to overfit. Might improve this if I find more time and get a better GPU.
landmark-detection-and-tracking-udacity
Udacity project. Implements SLAM (Simultaneous Localisation and Mapping) for an entity in a 2D world with various landmarks.
SpeechTextValidation
B.Sc. Dissertation Code. Compare Maltese speech to the prompt that was read out to ensure they match. Generate syllable distance and segment distance between the speech and text. Unsupervised and low-resource.
tv-script-generation-udacity
Udacity project. Trains an RNN on 9 seasons of Seinfeld scripts and generates new scripts.