2D-VQ-AE-2

2D Vector-Quantized Auto-Encoder for compression of Whole-Slide Images in Histopathology

How to run

Installation

See INSTALL.md.
We use PDM as python package manager (https://github.com/pdm-project/pdm).

Locally

set CAMELYON16_PATH and run train.py:

CAMELYON16_PATH=<camelyon-path> python train.py

Lisa

set CAMELYON16_PATH, and append --multirun to automatically submit a sbatch job through submitit.

• If CAMELYON16_PATH is a folder, the dataloader loads the dataset over the network.
• If CAMELYON16_PATH is a .tar, the file is copied to $SCRATCH of the allocated node, and the dataset is loaded locally.

CAMELYON16_PATH=<camelyon-path> python train.py --multirun

Change node type by overwriting the node config, e.g.:

CAMELYON16_PATH=<camelyon-path> python train.py hydra/launcher/[email protected]=gpu_titanrtx --multirun

Results

Note on Mean Squared-Error results: input is channel-wise normalised to 0-mean, 1-std, using the following values, based on 10k patches:

	Red	Green	Blue
Mean	0.7279	0.5955	0.7762
Standard Deviation	0.2419	0.3083	0.1741

Top: original, bottom: reconstruction.
Input dimensionality: 256×256×[email protected]μm ordinal 8-bit, latent dimensionality: 32×32@16μm categorical 8-bit (i.e. 99.47% compression), 0.900 MSE.

Input dimensionality: 512×512×[email protected]μm ordinal 8-bit, latent dimensionality: 32×32@16μm categorical 8-bit (i.e. 99.87% compression), 0.800 MSE.

Research

If this repository has helped you in your research we would value to be acknowledged in your publication.

Acknowledgement

This project has received funding from the European Union’s Horizon 2020 research and innovation programme under grant agreement No 825292. This project is better known as the ExaMode project. The objectives of the ExaMode project are:

1. Weakly-supervised knowledge discovery for exascale medical data.
2. Develop extreme scale analytic tools for heterogeneous exascale multimodal and multimedia data.
3. Healthcare & industry decision-making adoption of extreme-scale analysis and prediction tools.

For more information on the ExaMode project, please visit www.examode.eu.