MétaCan
Menu
Back to cohort

FTASD - A Fine Tuning Approach for Stable Diffusion Models

2024· article· en· W4404033230 on OpenAlex

Why this work is in the frame

A frame that forgets how it found something cannot be audited. These are the routes that admitted this work.

affAt least one author lists a Canadian institution in the pinned OpenAlex snapshot.

Bibliographic record

Venuenot available
Typearticle
Languageen
FieldComputer Science
TopicAdvanced Mathematical Modeling in Engineering
Canadian institutionsUniversity of Calgary
Fundersnot available
KeywordsDiffusionComputer sciencePhysicsThermodynamics

Abstract

fetched live from OpenAlex

Image generation is one of the critical tasks performed in today’s world for the expansion of research domains like computer vision and Generative Artificial Intelligence (Generative AI). Therefore, there is an ultimate need for 2-D image synthesis models, which are already introduced by the researchers in the form of stable diffusion models. Recently, various companies like OpenAI and Google AI introduced such models in the computer vision industry. The fundamental approach is to generate target images through a diffusion process. The various applications of stable diffusion include text-to-image generation, image restoration, image-to-image generation, video generation, and facial restoration. In recent years, multiple development plans have been incorporated for improving the image generation models. The improvements have been done in the form of improving the loss function, architecture designs and optimization method. In this paper, we propose a fine-tuning method for improving the performance of various image generation models using Stable Diffusion (SD). Our major focus is on various stable diffusion models which involve text (prompt) to image generation methodology for generating various synthetic images. In our fine tuning process, we leveraged the KerasCV and trained the pre-trained Stable Diffusion Model on a diversified POKEMON (BLIP caption generated) dataset fetched from the Hugging Face database. Our model outperformed the existing KerasCV stable diffusion model which is responsible for text to image generation. We also fine-tuned the model using our self collected Keji National Forest dataset and it produced outstanding application specific results.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.

Full frame distilled prediction

Teacher imitation

Not calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.

metaresearch head score (Codex)0.000
metaresearch head score (Gemma)0.000
Version: codex-gemma-dda1882f352aValidation status: machine_predicted_unvalidated
Candidate categoriesnone
Consensus categoriesnone
DomainCandidate signal: none · Consensus signal: none
Study designCandidate signal: Simulation or modeling · Consensus signal: none
GenreCandidate signal: Methods · Consensus signal: Methods
Teacher disagreement score0.377
Threshold uncertainty score0.323

Codex and Gemma teacher scores by category

CategoryCodexGemma
Metaresearch0.0000.000
Meta-epidemiology (narrow)0.0000.000
Meta-epidemiology (broad)0.0000.000
Bibliometrics0.0000.000
Science and technology studies0.0000.000
Scholarly communication0.0000.001
Open science0.0000.000
Research integrity0.0000.000
Insufficient payload (model declined to judge)0.0000.000

Machine scores (provisional)

The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.

Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.

Opus teacher head0.039
GPT teacher head0.260
Teacher spread0.221 · how far apart the two teachers sit on this one work
Validation statusscore_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from it