https://github.com/daniel-e/tinyimages
Tools for the Tiny Images database
https://github.com/daniel-e/tinyimages
Last synced: 3 months ago
JSON representation
Tools for the Tiny Images database
- Host: GitHub
- URL: https://github.com/daniel-e/tinyimages
- Owner: daniel-e
- Created: 2016-01-05T12:57:06.000Z (about 10 years ago)
- Default Branch: master
- Last Pushed: 2016-01-26T18:33:33.000Z (about 10 years ago)
- Last Synced: 2023-10-29T18:43:16.441Z (over 2 years ago)
- Language: Python
- Size: 23.6 MB
- Stars: 1
- Watchers: 2
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# User Stories
- [x] Search the nearest neighbours of the images stored in data/images/
- [x] on raw pixel data
- no image processing / feature scaling is done
- no dimension reduciton is done
- the Euclidean distance is computed on the raw pixel data
- [ ] after applying the Sobel operator to the raw pixel data
- [ ] on tiny images dataset
- [x] on toy data
- [ ] describe results
- green pixel probelm
- [ ] after applying dimension reduction via PCA
- [ ] on raw pixel data
- [ ] after applying the Sobel operator to the raw pixel data
# Tasks
- [ ] dimension reduction
- [x] compute mean
- [x] compute standard deviation
- [x] compute covariance matrix
- [ ] compute pca
- [ ] plot S (from svd)
- [x] image processing
- [x] Sobel operator
- [x] implement the Sobel operator
- [x] include the Sobel operator as a filter in the knn search
- [ ] visualize the filtered images in the knn summary report
- [x] query image
- [ ] query results
- [ ] ensure that the filtered image that is shown really has been used by knn
- [ ] solve the green pixel problem
- possible solutions
- [x] user another weighting scheme, e.g. 0.3, 0.3, 0.3
- does not work for toy dataset
- no edges are found anymore (even on edges that are clearly visible)
- [ ] compute the Sobel operator on each color channel
# Subtasks
# Changelog
- [x] clarify: why are there no edges for green lines after applying sobel
- lines do disappear after image is converted into grayscale
- the formular is: 0.21r + 0.72g + 0.07b
- the green component is quite near to the noise -> green pixel problem
- [x] show filtered query image in the summary report
- [x] knn.py can write the filtered query image into a file
- [x] integrated sobel from scipy
- [x] improved performance of gen.py to generate toy dataset
- [x] compute knn with sobel filter enabled
- [x] created library imageprocessing.py for reading/writing images