Game Playing Agent(Python)-
- Built a Game playing agent for Isolation game using different search algorithms and custom heuristic methods.
- Implemented alpha-beta pruning, mini-max and iterative deepening search algorithms to search the game tree.
- Using different evaluation functions, custom heuristics are developed to win the game in best possible method
- Analyzed and documented the performance using win-rate for each heuristic against different game starting positions.(Random,Center,Cornor etc.)
Dog Breed Classifier(Python,TensorFlow,Keras,OpenCV,Scikit-learn)-
- Built a Classifier using Machine Learning classifier-Convolutional Neural Networks(CNNs),TensorFlow,Keras that classifies the dog breed correctly if a dog image is given. Or if a human image is given, it gives the similar dog breed.
- Dataset of dog images are divided into test,train and validation set to test the problem of over-fitting and under-fitting of the models
- Used ResNet-50 model which is pretrained on the ImageNet dataset to detect dog images and used OpenCV’s Haar feature-based cascade classifiers to detect human faces
- Trained the CNN using transfer learning using pre-trained VGG-19,ResNet-50,Inception,Xception models to reduce training time.
Time Series Prediction and Text Generation(Python,TensorFlow,Keras):
- Built a model that predicts the stock prices of a company and generates sentences based on a novel using Machine Learning classifier- Recurrent Neural Networks(RNN) that generate sequences based on input data
- Used Long Short Term Memory(LSTM) cells architecture in the RNN.
- One-hot encoded the input characters
- The model predicts the next sequence of characters taking previous set of characters as input thus producing predicted sentences.
- Used Categorical Crossentropy loss and Keras RMSProp optimizer to compile the RNN model.
Computer Vision Capstone Project (Python,OPenCV,CNN,TensorFlow,Keras):
- Built a facial keypoint generator using deep learning concepts
- Used Haar Cascade Classifier to detect faces in the input image data.
- Detected edges of an image using OpenCV’s Canny edge detector
- Created and trained Convolutional Neural Network (CNN) to detect facial keypoints.
https://github.com/JessicaClaire/aind