Software Engineering Courses (SECourses)•3y ago

something I'm working on ATM. Looking good so far.

PPflanzen die Tanzen something I'm working on ATM. Looking good so far.

Furkan Gözükara SECourses•12/17/23, 8:32 PM

nice

FFurkan Gözükara SECourses nice

Pflanzen die TanzenOP•12/17/23, 9:04 PM

this is my first illustration character in SDXL. It's a lot easier than training a real person. Also a lot easier to put into context, even though all training images had white background.

rafraf•12/17/23, 9:05 PM

Hey @Dr. Furkan Gözükara i am running your autocropper_v7 but no cropping is taking place

rafraf•12/17/23, 9:06 PM

cropper-2.py3.62KB

rafraf•12/17/23, 9:06 PM

config.py629B

rafraf•12/17/23, 9:07 PM

here's my config for the variables (normally in the cropper_v7.py script. I am getting the full output to different ar folders, but the images are not actually cropped...

rafraf•12/17/23, 9:08 PM

rafraf•12/17/23, 9:08 PM

but the images in subfolders are all exactly like original portraits

rafraf•12/17/23, 9:08 PM

original image

rafraf•12/17/23, 9:08 PM

1024_1024 crop folder

rafraf•12/17/23, 9:09 PM

i am on visual studio copilot. not giving me a crop atm

rafraf•12/17/23, 9:10 PM

rafraf•12/17/23, 9:20 PM

ok well i can see some minor cropping, just i guess is due to model trying to preserve areas around face like neck. So i suppose i have to still run some manual cropping at Topaz stage of upscaling or with Birme. unless there is a threshold slider to force the model to crop

hozaggers•12/17/23, 9:31 PM

https://www.patreon.com/posts/89213064 am I correct in thinking we generally dont need to amend any settings not provided in the JSON files here? For example I noticed using /workspace/24GB_TextEncoder.json I have 8 epoch with 40 repeating steps. I don't need to change anything just because I have higher training images right? say 50 or 100? the training steps seem to work out still according to the formula about 35k when I print the training command

FFurkan Gözükara SECourses they cancelled me too haha

vp•12/17/23, 9:49 PM

careful if youtube retaliates in a similar fashion

rafraf•12/17/23, 9:57 PM

i really dont know what cropper does. its definitely not cropping to the aspect ration it says it is doing

Pflanzen die TanzenOP•12/17/23, 10:07 PM

Epoch 07 - 10 all score very good at weights from 0.7 - 1.0

Pflanzen die TanzenOP•12/17/23, 10:19 PM

Test from Epoch 09 at 0.85 weight. Pretty nice for a one shot with no negative prompt, just true to what can be seen in the grid prepended by Mike the boy and his dog WillyMike the boy and his dog Willy (Hah, just noticed a typo in my prompt)

rafraf•12/17/23, 10:38 PM

but i am guessing that is the limitation of the script. for full lenght images, one would have to pick the resizes from the 1024x1536 folder, which is giving me correct crops.

Rrafraf Hey @Dr. Furkan Gözükara i am running your autocropper_v7 but no cropping is tak...

Furkan Gözükara SECourses•12/17/23, 11:45 PM

it wont crop subject

Furkan Gözükara SECourses•12/17/23, 11:45 PM

then you need to run resizer

Furkan Gözükara SECourses•12/17/23, 11:45 PM

the resizer will focus face

Furkan Gözükara SECourses•12/17/23, 11:45 PM

and then it will crop with exact resolution

Vvp careful if youtube retaliates in a similar fashion

Furkan Gözükara SECourses•12/17/23, 11:46 PM

probably but i think youtube dont care out of youtube

Furkan Gözükara SECourses•12/17/23, 11:49 PM

i used auto cropper to crop chairs haha

Furkan Gözükara SECourses•12/18/23, 12:18 AM

i am gonna test triton on windows now

Furkan Gözükara SECourses•12/18/23, 12:19 AM

lets see if any speed gain

FFurkan Gözükara SECourses and then it will crop with exact resolution

rafraf•12/18/23, 12:41 AM

Yes it is working. Too bad the script doesn’t apply different aspect ratio crops according to his long the image is

AIGambino•12/18/23, 12:51 AM

I do a test render using ddim sampling method, 30 steps.. It looks like this, which I am ok with. But when I select my animatediff, it turns to this dark red backdrop. I keep the seed the same so no idea.. Does this have something to do with controlnet?

AIGambino•12/18/23, 12:52 AM

AAIGambino I do a test render using ddim sampling method, 30 steps.. It looks like this, wh...

Furkan Gözükara SECourses•12/18/23, 1:06 AM

i dont have experience with animateddiff yet

Stefanoid9343•12/18/23, 4:32 AM

Good night group any of you knows how to create those talking avatars with videos? I would like to create those but with training images

goalpwr•12/18/23, 4:56 AM

@Dr. Furkan Gözükara trying to apply this video: https://www.youtube.com/watch?v=343I11mhnXs however when running the script I'm getting the following error

File "N:\AI\Stable_Diffusion\ClassImages\SECourses\Deepface\venv\lib\site-packages\tensorflow\python\tf2.py", line 21, in <module>
from tensorflow.python.platform import _pywrap_tf2
ImportError: DLL load failed while importing _pywrap_tf2: A dynamic link library (DLL) initialization routine failed.

I've already tried installing the zlibwapi.dll into the scripts folder. I have an older computer with an NVIDA 3060 any assitance would be greatly appreciated and thanks for all your very detailed explanations!

YouTubeSECourses

How To Find Best Stable Diffusion Generated Images By Using DeepFac...

If you are also getting tired of trying to find good images among thousands of generated images you don't have to anymore. By using #DeepFace AI library, you can sort images by their similarity to your target images and quickly find the best Stable Diffusion #DreamBooth LoRA trained model generated images. I am explaining everything step by step...

Lorelai AI•12/18/23, 5:10 AM

The old comfy install I had isn't booting up, I tried a fresh install and had the same issue. It's probably a dumb item. I hadn't started it since we changed over Python.

PPflanzen die Tanzen Test from Epoch 09 at 0.85 weight. Pretty nice for a one shot with no negative p...

Hustle with AI•12/18/23, 6:22 AM

Nice work! How are you going about maintaining the consistency in the character?

HHustle with AI Nice work! How are you going about maintaining the consistency in the character?

Pflanzen die TanzenOP•12/18/23, 7:19 AM

It’s already trained like this. I can now prompt Mike the boy with or without his dog Willy or just Willy the dog alone. The character stays consistent most of the time. Good enough to pick from a batch of 4 ATM. I will train at least one more version, but I’ll try dreambooth next and add a lot more images which I can now easily generate. The first 24 pages book (prototype) is already done.

PPflanzen die Tanzen It’s already trained like this. I can now prompt Mike the boy with or without hi...

Hustle with AI•12/18/23, 7:36 AM

Ok, I was doing something similar with Dreambooth but havent been able to do so. If you are following any tutorial or decide to make one or have any tips :D, pls do share.
I want to create a comic strip with few characters which can remain consistent across their poses, expressions etc. So looking for ideas around it!

HHustle with AI Ok, I was doing something similar with Dreambooth but havent been able to do so....

Pflanzen die TanzenOP•12/18/23, 8:41 AM

Yeah. I might create a tutorial about the process at some point. Not sure if video or blog or even both. . Also my training data and Lora will be open-sourced first research (no commercial use)

Pflanzen die TanzenOP•12/18/23, 9:38 AM

First prototype.

Mike_and_the_Castle_small.pdf3.68MB

Pflanzen die TanzenOP•12/18/23, 10:03 AM

My process allows creating new character sheets (I trained character sheets, solo images and coupled images.
Looks like it was extremely effective

Hustle with AI•12/18/23, 10:05 AM

Yeah looks great. So you trained this model on Dreambooth? Or is it Lora?

Pflanzen die TanzenOP•12/18/23, 10:10 AM

just a standard LoRA with my usual settings. It is v2.

version was trained on 14 images (including solo, coupled and character sheets) all on white background
version included 35 images with more poses (including solo, coupled and character sheets) all on white background

Next will be dreambooth, with ~ 60 - 100 images. Will have to test what works best.

All training images are 4k - 8k, and captioned in a very similar manner

Mike the boyMike the boy
Willy the dogWilly the dog
Mike the boy with his dog WillyMike the boy with his dog Willy
Mike the boy is runningMike the boy is running
Mike the boy with hands in pocket, upper bodyMike the boy with hands in pocket, upper body
Mike the boy with mouth open, screaming, close-upMike the boy with mouth open, screaming, close-up
.. and so on.

See the attached screenshot on repeats. (the caption is done via text files, so only the number is used from those folder names)

I trained 10 epochs of which 7 - 10 at 0.7 - 1.0 weight are actually usable.

Hustle with AI•12/18/23, 10:16 AM

got it, thanks for sharing

I am using img2img with Cheyenne model using Canny controlnet to convert my imgs to cartoon. It is looking good so far. But next step is to train models for more flexibility.

HHustle with AI got it, thanks for sharing 👍 I am using img2img with Cheyenne model using C...

Pflanzen die TanzenOP•12/18/23, 10:44 AM

I only trained on whimsical watercolor images (as seen in the results). When I use a photography style on them I get these (1-shot) which is not perfect, but still quite impressive, considering I never trained any photos

SStefanoid9343 Good night group any of you knows how to create those talking avatars with video...

Furkan Gözükara SECourses•12/18/23, 10:50 AM

i worked on a repo but its resolution too low

Furkan Gözükara SECourses•12/18/23, 10:50 AM

https://github.com/yoyo-nb/Thin-Plate-Spline-Motion-Model

GitHub

GitHub - yoyo-nb/Thin-Plate-Spline-Motion-Model: [CVPR 2022] Thin-P...

[CVPR 2022] Thin-Plate Spline Motion Model for Image Animation. - GitHub - yoyo-nb/Thin-Plate-Spline-Motion-Model: [CVPR 2022] Thin-Plate Spline Motion Model for Image Animation.

Ggoalpwr @Dr. Furkan Gözükara trying to apply this video: https://www.youtube.com/watch?...

Furkan Gözükara SECourses•12/18/23, 10:51 AM

hello welcome. you downloaded from our patreon right?

Furkan Gözükara SECourses•12/18/23, 10:51 AM

can you tell me what do you see when you open cmd and type python?

LLorelai AI The old comfy install I had isn't booting up, I tried a fresh install and had th...

Furkan Gözükara SECourses•12/18/23, 10:51 AM

you can activate venv and install scipy manually

Furkan Gözükara SECourses•12/18/23, 10:52 AM

i also have auto installer : https://www.patreon.com/posts/92013455

something I'm working on ATM. Looking good so far.

Similar Threads