Summary

This document contains the work described in the AI for Smart City Powerpoint presentation and lists various AI projects and their statuses. The areas in which we develop solutions are:

1. Computer vision
2. NLP
3. Tabular-based

Computer Vision

Data capturing pipelines

Traffic images

• This deals with capturing images from public traffic camera APIs.
• More information is found here: Image capturing pipeline

Image labelling pipelines

• Labelling pipeline for DNV Sewer videos
• Work done by Vannary and Pavan

Image classification

Note

There is a difference between multi-class and multi-label image classification.

• Multiclass classification means a classification task with more than two classes; e.g., classify a set of images of fruits which may be oranges, apples, or pears. Multiclass classification makes the assumption that each sample is assigned to one and only one label: a fruit can be either an apple or a pear but not both at the same time.
• Multilabel classification assigns to each sample a set of target labels. This can be thought of as predicting properties of a data-point that are not mutually exclusive, such as topics that are relevant for a document. A text might be about any of religion, politics, finance or education at the same time or none of these.

Binary

• Binary classification (Flood or no flood in images) -- Jupyter Notebook
• Binary classification (Model to find text in the center of the image) -- Jupyter notebook
  • Can be used for comparison with the Google Vision API

Multi-class

• Multi-class classification on the COCO dataset
• Multi-class classification on the Imagenet dataset (used a base model for transfer learning)

Multi-label

• Multi-label image classification for one million Sewer ML images.

Image segmentation

• Segmenting floods and other objects in the image
• Microsoft Azure: AI for humanitarian action - The trained model did binary segmentation with a high Jaccard Coefficient of 0.91.
• Flood

Object detection

• YOLO object detection results done using Ultralytics.

NLP

• Sentiment Analysis for Stock Market Prediction (sentimarket)
• Data science for Water Distribution Utility Network
• Presentation for Dr. Vered Shwartz at UBC.

Tabular

H2O AI Wildfire Challenge – Winner

• Wildfire prediction with gradient boosting
  • Deployed Web App -- Link
  • The app achieved two kinds of predictions
    • Fire event prediction
    • Fire size prediction

Timeseries forecasting

• Streamflow prediction and forecasting
• Bi-LSTM based models for time series forecasting

BSM1

• Multivariate multi-step time series forecasting using Deep learning and gradient boosting

Work with geospatial data

Building footprints

• USGS LiDAR data
• DEM → extraction of building footprint