search

mlmed / torchxrayvision

TorchXRayVision: A library of chest X-ray datasets and models. Classifiers, segmentation, and autoencoders.
https://mlmed.org/torchxrayvision
Apache License 2.0
936 stars 220 forks source link

Question about "all" model and results #139

Closed MohammedSB closed 1 year ago

MohammedSB commented 1 year ago

Hello,

So I know that the "all model" is trained on multiple datasets, but is it trained on the entirety of those datasets (all training examples) or just a subset? Also, I wanted to ask where I can find the results of this model on all of the test sets, including NIH. I am not sure how to interpret the benchmark results in the markdown, it says:

Below is an example of running different models on the nih dataset and writing the output as markdown.

But in the markdown, none of the titles for the tables is for NIH.

ieee8023 commented 1 year ago

It is trained on a subset (I think 80%) but the split is lost so consider it to be all samples.

It seems I didn't run the benchmarks on the NIH dataset. I'll try to get to that soon.

I fixed the typo which should have said PadChest instead of NIH. So that example runs those models on the PadChest dataset and this results in a row for the results of each model.

ieee8023 commented 1 year ago

I added benchmark results for the NIH dataset for 3 models. The chexpert model is running now but it might take a day or so.

MohammedSB commented 1 year ago

Thank you very much! I really only needed the results for the all model.