~/talk

SaFace: Towards Scenario-aware Face Recognition via Edge Computing System

on YouTube 11m
Talkat 3rd USENIX Workshop on Hot Topics in Edge Computing (HotEdge '20) · 2020-06-25..06-26· [machine-learning, computer-vision]

speakers

name:Zhe Zhou
links:none on file
name:Bingzhe Wu
links:none on file
name:Zheng Liang
links:none on file
name:Guangyu Sun
links:none on file
name:Chenren Xu
links:none on file
name:Guojie Luo
links:none on file

resources

video:https://youtu.be/BkGUM0MDadI
slides:none on file

summary

Prototype implements a three-layer edge system that pretrains a CNN in the cloud, runs detection/alignment/tracking on smart cameras, and does inference plus unsupervised online learning on edge servers by mining triplets from tracked faces to fine-tune models. Evaluation uses two surveillance video sequences, three FR models, and shows accuracy gains across all models; a 32% relative improvement in online-learning throughput comes from a context-aware scheduler, while partial fine-tuning speeds up training at the cost of accuracy.

AI-written summary · about

chapters (6)

  1. 0:00Introduction
  2. 1:03The recognition workflow
  3. 2:26Unsupervised online learning
  4. 4:58System architecture
  5. 6:59Experiments and results
  6. 9:31Future work and conclusion

related talks

~/search
type to search 6,249 conferences, 138,987 speakers and 185,172 talks
~/help
cmdctrl k
search every conference, speaker and series
/
focus the filter search
h
home
c
conferences
s
speakers
e
series
t
light / dark — the status bar button does the same
?
this help
esc
close this, then clear every filter — or just the search, while you are typing in it

Ignored while you are typing. Pages without a filter rail have no search to focus.