Spaces:
Sleeping
Sleeping
Commit History
-
UI: Increase webcam border thickness, viewport size, and adjust layout for better premium experience
hoda.fakhar01c8852
-
docs: add face verification screenshot to README
hoda.fakharf7b4320 -
FINAL FIX: Enable Face Verification weights and fix 0-byte image serving
hoda.fakhar67ac72b -
docs: update README with live URL and accurate tech details
hoda.fakhar- Add direct link to https://hodfa71-vision-ai-engine.hf.space - Add free-tier cold-start performance note - Update TensorFlow version to 2.15.0 - Remove outdated PIL sharpening mention and vgg_face_weights entry - Remove speculative hosting options section (app is live)26086fb -
docs: add direct live URL and free-tier performance note
hoda.fakhar77417bf -
fix: use tensorflow/tensorflow:2.15.0 base image to avoid 300MB pip download
hoda.fakhar5e0e5c7 -
debug: split pip steps with echo markers so build logs show progress
hoda.fakhar368c8bf -
UN-BRICK SPACE: Remove LFS trackers for images to fix hanging build
hoda.fakharad27889 -
Fix permission error: switch from /app to /home/user in build script
hoda.fakhara01d457 -
Optimize Hugging Face build context and permissions
hoda.fakhar5ff59cc -
debug: add progress-bar to tensorflow install to show download progress in logs
hoda.fakhar79134ab -
fix: use tensorflow-cpu==2.15.1 (2.13.0 has ABI incompatibility with Debian Bookworm)
hoda.fakhar53c0ef5 -
fix: split pip install into steps for visibility, isolate tensorflow install
hoda.fakharaae2413 -
fix: downgrade to tensorflow-cpu==2.13.0, add pip timeout/retry flags to prevent build hang
hoda.fakhar271c6f2 -
fix: pin tensorflow-cpu==2.15.1, add 5min download timeout to prevent build hang
hoda.fakhar7189f3f -
fix: remove aggressive sharpening, fix gender confidence mismatch in live view
hoda.fakhar84b1835 -
feat: enable retinaface detector, download retinaface.h5 from HF Hub during build
hoda.fakharbb776bc -
fix: replace libgl1-mesa-glx with libgl1 for Debian Trixie
hoda.fakhar779b029 -
fix: remove retinaface from live view, clean unused import
hoda.fakharcdfeda4 -
fix: stop 500 crash + reliable model download from HF Hub
hoda.fakhar- ml_engine.py: fallback result now includes gender/race/emotion keys so analysis_result.html template never raises KeyError -> no more 500 - recognition_routes.py: use shared analyzer singleton (removes duplicate FaceAnalyzer instance that wasted memory) - build_models.py: try Hodfa71/deepface-weights HF Hub first, fall back to GitHub; HF Hub is fast/reliable within HF infrastructureff31e2d -
fix: OOM fix - models baked into Docker image, mobile responsive, 1 worker
hoda.fakharab8cebe