kalilzera/DeepFakes

fix: rename env vars for HF Spaces compatibility

feat: deepfake+genai integration + video support

Apply security patches: rate limiting, payload validation, and JSON error handling

fix(api): fix file upload race condition and add explicit unidentifiable file error handling

Fix 500 Error: Handle NoneType for pca.transform

fix(ui): Change image preview from object-cover to object-contain

Deploy Lovable React frontend + 6-class Gemini model via Docker Space

feat: replace Gradio with Docker SDK + Flask serving Lovable React frontend

docs: update confusion matrix showing SMOTE balance improvements

feat: retrain MLP classifier with SMOTE/Random OverSampling to fix Midjourney false negatives and balance 'Real' bias

fix: map integer classes to human readable AI model names in Gradio output

docs: add confusion matrix for the 5-class model to HF Spaces repository

docs: sync HF space metadata with the new Multi-Class classification info

feat: upgraded to a 5-class Multi-Class Identification Model distinguishing between Human, GAN, ChatGPT, Midjourney, and Stable Diffusion

feat: updated model weights with Defactify dataset to detect modern Diffusion AIs (Midjourney/DALL-E)

fix: double script tag, use @gradio/client to fix API routing

fix: keep gr.Image interactive with CSS hiding instead of visible=False

feat: full custom glassmorphism UI with drag-and-drop, scan animation, live API

fix: move Google Fonts to link tag, fix @import CSS restriction

feat: custom glassmorphism UI with gr.Blocks

fix: pin gradio==5.23.1 to fix audioop/pydub Python 3.13 incompatibility

feat: DenseNet121 + MLP deepfake detector (97.19% acc)

initial commit