quang/ai
VI Hire me
Computer VisionPaper

Face-Aware MAE

Identity-aware masked autoencoding: instead of masking random blocks, point MAE pre-training at the eyes, nose, mouth and central facial structure — and change nothing else.

identity-critical reconstruction error
−17.5%
added parameters
0
masking strategies compared
3
CelebA split
90/10

Idea

Standard MAE masking treats every patch equally. For faces, identity lives in a few regions. This study asks whether steering the masking policy toward those regions produces better face representations — without touching the encoder–decoder.

Controlled comparison

  • Random, block-wise and face-aware (landmark-guided) masking under one CelebA setup (90/10 split).
  • Evaluated with validation loss, localized reconstruction error, ablations, linear probing and zero-shot evaluation.

Result

Face-aware masking reduced identity-critical reconstruction error by 17.5% against block-wise masking while adding zero model parameters — only the masking policy changed.

→ More work

Other case studies

BTS Digital Twin — Viettel AI Race 2026

Novel view synthesis of telecom (BTS) towers from drone captures with 3D Gaussian Splatting — then large urban scenes under an inference-time limit on an NVIDIA H200 in the final round.

 iPBMS — real-time patient monitoring

Fall and seizure detection on live RTSP hospital camera streams — from pose model to on-premise production on two cameras, with alerts to care staff in under two seconds.

 Inference-Efficient ConvMAE

Can a masked-autoencoder backbone be cheaper at inference without losing accuracy? Four hybrid designs, one fair protocol, measured under TensorRT rather than guessed from parameter counts.

 Vietnamese License Plate Recognition Benchmark

A reproducible benchmark for Vietnamese automatic license plate recognition: three YOLO generations, three OCR pipelines, real RTSP traffic-camera data.
esc