Training data consultation

[yihp]

hi I found that in the mimic dataset, one report often corresponds to multiple chest X-rays. I would like to ask whether your research uses single image input or multiple image input?

[anicolson]

Hi @yihp, This work uses single-image, whereas our later work uses multi-image.

[yihp]

I just mentioned concat multiple image inputs this work, and I would like to ask if your subsequent work has been published? Can I know your thinking?

[anicolson]

Yes, pass each image for a report through the image encoder, then concatenate the outputs from the encoder. This is the standard way.

Get Outlook for iOShttps://aka.ms/o0ukef

---

From: yihp @.> Sent: Wednesday, June 26, 2024 7:04:21 PM To: aehrc/cvt2distilgpt2 @.> Cc: Nicolson, Aaron (H&B, Herston) @.>; Comment @.> Subject: Re: [aehrc/cvt2distilgpt2] Training data consultation (Issue #17)

I just mentioned concat multiple image inputs this work, and I would like to ask if your subsequent work has been published? Can I know your thinking?

— Reply to this email directly, view it on GitHubhttps://github.com/aehrc/cvt2distilgpt2/issues/17#issuecomment-2191189115, or unsubscribehttps://github.com/notifications/unsubscribe-auth/AGHGZ7VHH4FZWDVRLGO37WLZJJ7ZLAVCNFSM6AAAAABJ5NSYESVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDCOJRGE4DSMJRGU. You are receiving this because you commented.Message ID: @.***>

[yihp]

Yes, pass each image for a report through the image encoder, then concatenate the outputs from the encoder. This is the standard way. Get Outlook for iOShttps://aka.ms/o0ukef … ____ From: yihp @.> Sent: Wednesday, June 26, 2024 7:04:21 PM To: aehrc/cvt2distilgpt2 @.> Cc: Nicolson, Aaron (H&B, Herston) @.>; Comment @.> Subject: Re: [aehrc/cvt2distilgpt2] Training data consultation (Issue #17) I just mentioned concat multiple image inputs this work, and I would like to ask if your subsequent work has been published? Can I know your thinking? — Reply to this email directly, view it on GitHub<#17 (comment)>, or unsubscribehttps://github.com/notifications/unsubscribe-auth/AGHGZ7VHH4FZWDVRLGO37WLZJJ7ZLAVCNFSM6AAAAABJ5NSYESVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDCOJRGE4DSMJRGU. You are receiving this because you commented.Message ID: @.***>

Is this your article: "Longitudinal Data and a Semantic Similarity Reward for Chest X-ray Report Generation"

[yihp]

Hi @yihp, This work uses single-image, whereas our later work uses multi-image. Hi

Could you please confirm if your subsequent work is titled "Longitudinal Data and a Semantic Similarity Reward for Chest X-ray Report Generation"?

[anicolson]

Yes, that is our work.

This is our latest work:

https://arxiv.org/pdf/2406.13181

[yihp]

Yes, that is our work.

This is our latest work:

https://arxiv.org/pdf/2406.13181

thank you!