This is a compilation of YouTube videos that has been processed for use in a machine learning dataset. The goal is to facilitate the creation of tools that provide per-pixel labels for features ranging from the surgeon's tools to the patient's tissues. I think such tools have the potential to greatly enhance the value of surgical e-Shadowing, giving students the ability to click parts of the video itself to get information about the anatomy under the cursor. This feature would go well with e-Shadowing Transcriber, which automates the retrieval of supplementary materials.
- label more frames
- overfit model on toy dataset
- create a frame-filter model using t4vd
- train a variational auto-encoder on the t4vd-filtered frames
- determine if the VAE improves data efficiency
- evaluate amazon mturk with this project's data
- e-Shadowing Transcriber: an application for enhancing the medical e-Shadowing experience
- thavlik portfolio: my showcase of medical software projects, of which this repository is a part
Video data is property of the respective authors:
All code in this repository is released under MIT / Apache 2.0 dual license, which is extremely permissive. Please open an issue if somehow these terms are insufficient.