https://github.com/joshuacwnewton/swift-classification
A sister repository to "swiftwatcher" for testing classification between images of chimney swifts and images of other bird species.
https://github.com/joshuacwnewton/swift-classification
birds chimney-swifts deep-learning machine-learning motion-detection motion-tracking neural-network python swift-classification
Last synced: 3 months ago
JSON representation
A sister repository to "swiftwatcher" for testing classification between images of chimney swifts and images of other bird species.
- Host: GitHub
- URL: https://github.com/joshuacwnewton/swift-classification
- Owner: joshuacwnewton
- Created: 2020-01-22T22:31:47.000Z (over 6 years ago)
- Default Branch: master
- Last Pushed: 2020-08-09T17:51:37.000Z (almost 6 years ago)
- Last Synced: 2025-02-16T00:25:34.330Z (over 1 year ago)
- Topics: birds, chimney-swifts, deep-learning, machine-learning, motion-detection, motion-tracking, neural-network, python, swift-classification
- Language: Python
- Homepage:
- Size: 992 KB
- Stars: 0
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
> **Please note**: the development of this project was interrupted due to the unprecedented circumstances of COVID-19. The intermediate results of this research are provided for transparency. As this was a project for an undergraduate Honours Thesis, development has ceased, and the author has moved onto future projects.
### swift-classification
> A sister project to the previous SwiftWatcher collaboration between the University of Victoria's Computer Vision Lab and Algoma SwiftWatch. Completed by Joshua Newton as part of Honours Thesis during Spring 2020 term.
The `swiftwatcher` project previously aimed to **detect, track, and analyze the motion** of chimney swifts in video footage. However, video footage typically contains the motion of more than just chimney swifts. _(For example, seagulls, crows, and other birds.)_ So, to prevent non-swift birds from being tracked, it would be useful to distinguish between chimney swifts and other types of birds.
Images will be extracted from available source videos using the existing `swiftwatcher` project. Then, these images will be separately curated into a new dataset using the tools created within this `swift-classification` repository. This new dataset will then be used to train and evaluate a machine learning classifier in an attempt to filter out non-swift motion. Once trained and evaluated, the model will then be integrated into the `swiftwatcher` project for inference purposes.
### Approach
~~This approach is currently undergoing active development.~~ Preliminary proposed methods include:
1. Haar-like features + AdaBoost classifier
2. HoG features + SVM classifier
3. Convolutional Neural Network-based classifiers
These approaches will be explored in greater detail over the coming months.
### Further Reading
A presentation slide deck has been created, which provides greater detail into the problem domain and available datasets. The slide deck can be viewed using this link: [Chimney Swift Classification Overview](assets/Chimney-Swift-Classification-Overview.pdf)