Leveraging Text-Conditioned Diffusion in the Latent Space for Synthetic Cancer Pathology Image Generation

Capstine Thesis Submission

[CS-4999-01] Ashoka University

This is the repository attached to CS-4999-01, my submission for the Computer Science ASP Capstone Thesis, 2024. This repository contains 5 folders

1. FID-calculator
2. GPT-API
3. PathLDM
4. TCGA_stats
5. latent-diffusion

FID-calculator

This sub-directory corresponds to the standardised synthetic image generation, and model evaluation workflow.

GPT-API

This sub-directory corresponds to the created summary generation pipelines.

PathLDM

This is an adaptation of the original repository of PathLDM, this consists of edits that have been made to the original code in order to better ensure the workability of the code.

TCGA_stats

This sub-directory consists of the code used to generate the statistics of the different data present in the TCGA-BRCA dataset.

latent-diffusion

THis is the sub-directory corresponding to the original latent diffusion model, which serves as the inspiration of PathLDM, and our work.

Further details are provided in the individual sub-directories.