Projects

Computer Vision

Emotional Response

Python, CNN, Keras, Deep Learning, Image Classification

Trained a CNN on artwork to classify the emotional response a viewer would feel, achieving better performance after simplifying to binary happy/sad classification.

Code

Deblurring Computer Vision

Python, Optical Flow, Deep Learning, Computer Vision

Implemented and improved a video deblurring research paper by using optical flow thresholding to eliminate artifacts in near-static regions.

Visual Search

Python, VGG16, ImageNet, K-Means Clustering, PyTorch

Grouped similar products by visual appearance using VGG16 embeddings and clustering, resolving mirroring and grayscale issues via spatial pooling.

Variable Multiple Instance Learning

Python, PyTorch, CNN, Deep Learning

Explored variable-input fixed-output learning for 3D object modeling, implementing a CNN baseline and proposing an RNN-inspired embedding approach.

Nutrition Table Mapping

Python, Deep Learning, Neural Networks, Synthetic Data Generation

Digitized nutrition tables from product images end-to-end using a neural network trained on synthetically generated bounding box data.

Logo Prediction

Python, OpenCV, Keypoint Matching, Deep Learning

Identified company logos on product images using keypoint matching and a dynamically generated logo collage for confident brand attribution.

Fashion Design Overlay

Python, OpenCV, Computer Vision

Developed a system to overlay arbitrary designs onto clothing worn by models by mapping intersection points of a checkered pattern to simulate cloth curvature.

Cropping Tool and Object Tracking

Python, OpenCV, Deep Learning, Computer Vision

Built a dataset generation tool that crops a product once and then tracks it through video frames as it moves and transforms.

Colour Based Item Selection

Python, OpenCV, Optical Flow, Machine Learning

Identified products placed in a cart by matching dominant colors from the camera feed against a product color database using optical flow and skin detection.

Detecting Products Put Into a Cart

Python, Computer Vision, Deep Learning, OpenCV

Detected products being added to a cart by sequentially analyzing hand and hand-with-product detections from a camera feed.

Machine Learning

Prompt Engineering

Python, ChatGPT, GPT-3.5, Prompt Engineering

Used ChatGPT and prompt engineering to summarize customer support conversations and speed up escalations at Dell.

Automated Data Analysis

Python, Machine Learning, Pandas, scikit-learn

Built a meta-language and ML framework enabling non-technical users to extract insights from company data without a data analyst.

Reinforcement Learning

Value Smoothing via Latent Embedding Similarity

Python, PyTorch, Reinforcement Learning, Embeddings

Modified RL reward structure using embedding-space similarity to improve learning in sparse-reward environments, experimentally outperforming standard algorithms.

Code