My app has a simple GUI in windows. 1. Accesses jpeg files on local PC. 2. Uploads hand-drawn images. 3. Interfaces with GPT4 image option & prompted to compare the jpegs. 4. The app displays the outcome of the comparison in the GUI window.
Budget: $750 – $1,500 AUD
The project is a desktop application running on windows 11.
Steps:
1. The application has a simple graphical user interface (think GPT4)
2. From the user’s computer the app will randomly select a pair of jpegs from two folders (A and B) each containing 128 folders (A1-A128 and B1-B128), each containing one jpeg.
(Folder path e.g. C:\Users\zonma\Documents\Parvarti\Pixtures\A\A42>A42.jpeg
and
C:\Users\zonma\Documents\Parvarti\Pixtures\B\B42<B42.jpeg).
The app notifies the user readiness to move to step 3. (e.g. Ready to upload drawing!)
3. The user uploads a jpeg of a hand-drawn image to the app.
4. The application interfaces with GPT4's default option (image upload) where it’s prompted to compare the drawing with one pair of images. (Prompts are in three steps within one chat session. Each step includes a picture upload. This step has been tested manually, see prompts below)
5. The application returns the outcome of the comparison by GPT4 to the app window. E.g., "The drawing "X" most closely matches picture “A42" or "B42". (The app must display only the label e.g. "A42". No picture and no verbal descriptions).
GPT4 - 3 step prompt.
Here is a picture with the label "A42". Please describe it.
Here is a picture with the label "B42". Please describe it.
Here is a hand drawn image with the label "X". Which does "X" most closely resemble, picture "A42" or "B42"?
Option to repeat the process with new hand drawn image but with the same randomly chosen pair as many times as user likes.
Steps:
1. The application has a simple graphical user interface (think GPT4)
2. From the user’s computer the app will randomly select a pair of jpegs from two folders (A and B) each containing 128 folders (A1-A128 and B1-B128), each containing one jpeg.
(Folder path e.g. C:\Users\zonma\Documents\Parvarti\Pixtures\A\A42>A42.jpeg
and
C:\Users\zonma\Documents\Parvarti\Pixtures\B\B42<B42.jpeg).
The app notifies the user readiness to move to step 3. (e.g. Ready to upload drawing!)
3. The user uploads a jpeg of a hand-drawn image to the app.
4. The application interfaces with GPT4's default option (image upload) where it’s prompted to compare the drawing with one pair of images. (Prompts are in three steps within one chat session. Each step includes a picture upload. This step has been tested manually, see prompts below)
5. The application returns the outcome of the comparison by GPT4 to the app window. E.g., "The drawing "X" most closely matches picture “A42" or "B42". (The app must display only the label e.g. "A42". No picture and no verbal descriptions).
GPT4 - 3 step prompt.
Here is a picture with the label "A42". Please describe it.
Here is a picture with the label "B42". Please describe it.
Here is a hand drawn image with the label "X". Which does "X" most closely resemble, picture "A42" or "B42"?
Option to repeat the process with new hand drawn image but with the same randomly chosen pair as many times as user likes.