1.FF ♀ Cyberpunk Style

1.1CivitAI LoRA
1.2LoRA FA test run
1.3.Lycoris Test Run
2.FF 𓃥 𓃠 𓃰 𓃱 𓃯 𓃭 𓃸 𓃵 Cyberpunk Warfare
2.1 Lycoris H100 run


3.FF ♂ Cyberpunk Style tba
Description
Trained with official CivitAI Lora trainer with 16 epoch set up.
FAQ
Comments (7)
What do you think is the advantage of including depth maps in the dataset? Are you not afraid that the LoRa will only produce black and white images?
Including depth maps in the dataset serves as an additional feature for those who might want to utilize it in further training or testing endeavors. In this particular instance, depth maps are not actively used. However, I would recommend employing depth maps when working with final models, as they can really help the learning around objects / smoke and other ... i am doing this since 2.1 and all models. ... with caution tho https://latex.id/wp-content/uploads/2023/10/Strip-2-scaled.webp
ps: do love your nike high heels .. nice work 🤟 🥃
@idle Thank you very much. Maybe this could help with my car models. I have the issue with the 'Shelby Mini Cobra' model that sometimes the people are in front of, not behind, the windshield. I will train the model again with 2-3 additional depth maps. Maybe it will work better then.
@idle You mean you actually train depth maps aka greyscale softened images with a colorful dataset? Or that was a split thing? Still don't get my head around this though.
@denrakeiw Try getting a guy riding a motorcycle, or even sitting on a motorcycle, or next to a motorcycle that isn't too small or too large. Talk about cherry picking. Even XL fails so much at this.
Do you just caption them as "depth map"? I ignored 2.1's use of depth maps because they were ridiculously low res for me and I didn't use 2.1 much. How are you making them? I find too many depth maps to not be real G2H. And sadly, I find that my 3D app (Cinema 4D) isn't great at making them... though that might be because I've only tried difficult scenes like a toroidal space station. You need more than 8 bit, but ControlNet doesn't like 16 or 24 bit files. I ended up faking them in C4D, then in Photoshop.
@parallelepipedon That's why most of my vehicle models consist of mini motorcycles or mini cars and are trained to depict characters much larger.
Details
Files
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.







