1. Yeah, I call and get the receptionist. I tell them about the software and get their interest (which is always seemingly high). They give me the email address of their plastic surgeon. I email the surgeon with a link to this page while explaining to them that I'm creating this software but want to validate the idea before putting more time into it. Nobody responds to my emails.
2. The photos are right there in the link, but if you're not able to view them, to me it usually generates pictures that are close to indistinguishable from actual photos. The breasts look like breasts (unless I'm cramming five on them onto one chest and make them shaped oddly... lol)
I just need more pictures. I also need to apply a different image-to-image translation where I put "static" over JUST the segmented breasts, so as to not remove tattoos, freckles, etc. I've seen an awesome paper that does this called OASIS, but they've not released code. I want the synthesized picture to be as close as the original as possible (with the obvious exception of larger breasts).
Still I need to build the actual pipeline instead of running the images through my segmentation network and my SPADE network AND I need to maybe programmatically increase the segmented "breast" category to be larger and maybe place the nipple in a more desirable location.
1. I think the website does your project a disservice instead of amplifying it. This is because it seems to focus on the cool engineering side of the project (really cool!). Instead, I think it should focus on the value it could generate for the clinics and how close you are to make it work.
For example, you change the first paragraph as follows. (English is not my first language, so phrasing and grammar could certainly be improved.)
Current version:
"I'm trying to validate a program I've written to help your clients see what they might look like after a breast augmentation instead of using Photoshop or relying only on vests. Maybe they'd like to see how they could look by using Breastimate? Please take a look at what I've come up with so far and offer any feedback. (These results are from my rudimentary proof-of-concept)"
Suggestion:
The idea
Showing clients how they would look like after breast augmentation often requires Photoshop or wearing a vest. This costs time and money, and results are often inaccurate.
My goal is to immediately provide accurate results, reduce the costs to the clinics, and increase client satisfaction by using artificial intelligence and computer vision.
Proof-of-concept
Currently, my software simulates breast augmentation results that are quite close to real-life augmentation outcomes. The below pictures were produced in 30 seconds from uploading pre-augmentation photographs.
To improve the results and validate this approach, I am looking for breast augmentation clinics to partner in using and developing this software.
What I am offering:
- Free lifetime access to this software
What I am looking for:
- feedback on improving it from a clinician's perspective
- access to pre- and post-augmentation photographs to further increase the accuracy of produced results
2. I couldn't find the comparison pictures in the format that I described above.
3. I would move the technical details to the very bottom of the page or a separate page. Instead, to build trust, I would write a sentence or two about who you are, add your picture, and link to your resume or LinkedIn.
4. I would remove the "bizarre experiments" because they distract from the main idea and suggest your software is not working well. Someone who skims the text and doesn't know what labels or masks are may think that your software produces Total Recall-like breasts.
5. You could also address how your software is better / cheaper from other options. I found one example here: https://www.epsteinplasticsurgery.com/procedures/breast-proc... BTW, you need to make sure that the vast majority of clinics indeed relies on Photoshop and vests. Otherwise, your main selling point suggests that you don't know this space and may turn off people you are contacting.
Thank YOU for answering! I agree with all of your feedback, and the clarity of your restructured and prioritized text will definitely improve my chances of success. In the event of any sort of success, I'll be sure to return to share in the success.
I just need more pictures. I also need to apply a different image-to-image translation where I put "static" over JUST the segmented breasts, so as to not remove tattoos, freckles, etc. I've seen an awesome paper that does this called OASIS, but they've not released code. I want the synthesized picture to be as close as the original as possible (with the obvious exception of larger breasts).
Still I need to build the actual pipeline instead of running the images through my segmentation network and my SPADE network AND I need to maybe programmatically increase the segmented "breast" category to be larger and maybe place the nipple in a more desirable location.