visualizing_image_embeddings

Python implementation for visualizing lower-dimensional representations/embeddings of images that are created by first feature extraction using pretrained (ImageNet) ResNet-18 (default) and application t-distributed stochastic neighbor embedding (t-SNE) method on those ResNet-18 features

Installation

git clone [the url of visualizing_image_embeddings]
cd visualizing_image_embeddings
pip install -r requirements.txt

Usage

Tested on Python 3.8.10

python main.py -p [path of the input spreadsheet]

The data set, collection of textures in histological images of human colorectal cancer, by Kather et al. 2016 (https://zenodo.org/record/53169#.Y8U85nbMJNY) is used as sample input in this repository. Construction of the sample input spreadsheet is shown in "make_sample_spreadsheet.py" for the required format.

Sample generated plots

Thumbnail-plot

Dot-plot

License

MIT