update requirements.txt

fix(cleanup): final pre-release audit — emojis, unused vars, version, CI trigger, starlette CVE

Merge pull request #94 from abinaze/fix/deploy-bugs-v6.1

fix(api,deploy): correct AttributeError, missing lifespan, Dockerfile drift, XGBoost double-boost

Merge pull request #93 from abinaze/xgboost-trained-ensemble

feat(ensemble): — XGBoost trained classifier replaces hardcoded signal weights

fix(deps): constrain numpy<2.0.0 to maintain opencv-python compatibility

Merge pull request #92 from abinaze/fix/synax,bug

fix: audit pass — unused imports, None crash guard, emojis, accuracy claims, broken f-strings

Merge pull request #90 from abinaze/embedding-detector

fix: return None from load_model when no weights exist, return neutral 0.5 in detector

update requirements.txt

fix: update signal count to 26, fix perturbation non-determinism, fix DIRE import, patch starlette CVE

security: update vulnerable dependencies flagged by Dependabot

docs: document manifest.csv is generated locally

chore: remove manifest.csv from git tracking (82MB, use locally only)

feat: complete — trained EfficientNet detector, centroids, integrated into ensemble

Update index.html

Update README.md

Update README.md

Merge pull request #89 from abinaze/embedding-detector

docs: update README with theme descriptions and logo

design: three-theme frontend with animated Three.js, dark, light modes and real logo

feat: add EfficientNet training script — reads manifest.csv, trains binary classifier

Merge pull request #88 from abinaze/embedding-detector

fix: skip hash check on EfficientNet weights, fallback to random init

Merge pull request #87 from abinaze/embedding-detector

deps: add scikit-learn and tensorboard for Phase 2 embedding detector

feat: add OwnEmbeddingModel — EfficientNet-B0 architecture for AI detection

Merge pull request #85 from abinaze/dataset-foundation

fix: remove pinned python-magic-bin version that breaks Windows CI

data: add gitkeep for new real dataset folders

chore: split manifest into real/ai files, keep sample in manifest.csv

Merge pull request #84 from abinaze/dataset-foundation

fix: statistical detector test signal counts

chore: ignore parquet, tsv, and csv files except manifest

Merge pull request #83 from abinaze/dataset-foundation

feat: add dataset download scripts for real and AI sources

Merge pull request #81 from abinaze/dataset-foundation

docs: add full dataset registry — 8 real + 10 AI sources, 200K target

data: scaffold dataset directory structure with gitignore rules

Merge pull request #80 from abinaze/fix/all-issue

Merge branch 'main' into fix/all-issue

fix: correct signal counts in statistical detector tests

Merge pull request #79 from abinaze/feature/calibration-polish

Fix: relax tiny image confidence assertion — PRNU may return non-zero confidence

Fix: s[score] → s['score'] (NameError), fix cached_download → hf_hub_download

feat: Calibration, dedup fix, signal tests

Merge pull request #78 from abinaze/feature/dct-frequency

feat:DCT Frequency Artifact Detection (25th signal)