Hello! Here is some of my work with fine tuning diffusion models as a 1st year engineering student. All of the following are generated using SOTA open-sourced generation models that I fine tuned.
These images are all of me. I am not a narcissist lol, it is just that testing on yourself makes it easiest to scrutinize appearance and similarity.
VTON and Posing
The full walkthrough and example on the app can be seen in a video
Me in clothes I can not afford
Some progress
Start (October)
November
December
Results for avatar building
The full walkthrough and example on the app can be seen in a video
Garry Tan!!!!!??
-->
+
This is a body picture of my brother who is more athletic and better looking than me but I don't have any photos of myself
+
-->
Lighting details on this one needs more work. I am working to fix this by training LoRAs longer and on better data and preprocessing photos. Results are best when photos are taken from Iphone which will likely almost always be the case with future users. I just couldn't find neutrally lit photos of this person.
+
-->
Results for object-scene placement (Photos taken on an Iphone 11)
-->
-->
-->
