accuracy   1246

« earlier    

Face Recognition Software’s Capabilities failed 20x less in 2018 vs. 2014
"The test—performed in the 2010, 2014 and 2018 evaluations—judged how well an algorithm could match a person’s photo with a different one of the same person stored in a large database. This type of “one to many” search is often employed to check for a person who might be applying for a visa or driver’s license under a name different than their own.

The team found that just 0.2 percent of searches failed this year, compared with a 4 percent failure rate in 2014 and 5 percent in 2010. Failure means that when an image of a person’s face is submitted to the recognition software, it fails to return the matching face image that resides in the database.

All of the top-performing algorithms from the latest round make use of machine-learning software architectures called convolutional neural networks. According to NIST’s Patrick Grother, one of the report’s authors, the rapid advance of machine-learning tools has effectively revolutionized the industry."

The test—performed in the 2010, 2014 and 2018 evaluations—judged how well an algorithm could match a person’s photo with a different one of the same person stored in a large database. This type of “one to many” search is often employed to check for a person who might be applying for a visa or driver’s license under a name different than their own.

The team found that just 0.2 percent of searches failed this year, compared with a 4 percent failure rate in 2014 and 5 percent in 2010. Failure means that when an image of a person’s face is submitted to the recognition software, it fails to return the matching face image that resides in the database.

All of the top-performing algorithms from the latest round make use of machine-learning software architectures called convolutional neural networks. According to NIST’s Patrick Grother, one of the report’s authors, the rapid advance of machine-learning tools has effectively revolutionized the industry.
face  recognition  accuracy 
14 days ago by dandv
classification - ImageNet: what is top-1 and top-5 error rate? - Cross Validated
Now, in the case of top-1 score, you check if the top class (the one having the highest probability) is the same as the target label.

In the case of top-5 score, you check if the target label is one of your top 5 predictions (the 5 ones with the highest probabilities).
nibble  q-n-a  overflow  machine-learning  deep-learning  metrics  comparison  ranking  top-n  classification  computer-vision  benchmarks  dataset  accuracy  error  jargon 
21 days ago by nhaliday
[no title]
verification checklist for photos
photos  verification  checklist  accuracy 
24 days ago by dmensing
[no title]
Checklist for assessing accuracy of a video
accuracy  video  fake 
24 days ago by dmensing
What every computer scientist should know about floating-point arithmetic
Floating-point arithmetic is considered as esoteric subject by many people. This is rather surprising, because floating-point is ubiquitous in computer systems: Almost every language has a floating-point datatype; computers from PCs to supercomputers have floating-point accelerators; most compilers will be called upon to compile floating-point algorithms from time to time; and virtually every operating system must respond to floating-point exceptions such as overflow. This paper presents a tutorial on the aspects of floating-point that have a direct impact on designers of computer systems. It begins with background on floating-point representation and rounding error, continues with a discussion of the IEEE floating point standard, and concludes with examples of how computer system builders can better support floating point.
nibble  pdf  papers  programming  systems  numerics  nitty-gritty  intricacy  approximation  accuracy  types  sci-comp 
29 days ago by nhaliday
[1803.00085] Chinese Text in the Wild
We introduce Chinese Text in the Wild, a very large dataset of Chinese text in street view images.

...

We give baseline results using several state-of-the-art networks, including AlexNet, OverFeat, Google Inception and ResNet for character recognition, and YOLOv2 for character detection in images. Overall Google Inception has the best performance on recognition with 80.5% top-1 accuracy, while YOLOv2 achieves an mAP of 71.0% on detection. Dataset, source code and trained models will all be publicly available on the website.
nibble  pdf  papers  preprint  machine-learning  deep-learning  deepgoog  state-of-art  china  asia  writing  language  dataset  error  accuracy  computer-vision  pic  ocr 
4 weeks ago by nhaliday
Basic Error Rates
This page describes human error rates in a variety of contexts.

Most of the error rates are for mechanical errors. A good general figure for mechanical error rates appears to be about 0.5%.

Of course the denominator differs across studies. However only fairly simple actions are used in the denominator.

The Klemmer and Snyder study shows that much lower error rates are possible--in this case for people whose job consisted almost entirely of data entry.

The error rate for more complex logic errors is about 5%, based primarily on data on other pages, especially the program development page.
org:junk  list  links  objektbuch  data  database  error  accuracy  human-ml  machine-learning  ai  pro-rata  metrics  automation  benchmarks  marginal  nlp  language  density  writing  dataviz  meta:reading  speedometer 
4 weeks ago by nhaliday

« earlier    

related tags

000  107  2018-04-18-0  2018-04-18  2018  2019  2drms  3d  3dprinter  3dprinting  5  754  absence  accurate  adblock  aesthetics  aggregator  ai-control  ai  algorithm  algorithms  alternative  analytical-holistic  and  annotation  anthropomorphizing  api  apple  applications  approximation  apps  ar-15  archer  architecture  archives  arduino  arms  arrow  art  article  articles  artificial  artificialintelligence  asia  assembly  audio  authentication  authenticity  authority  automation  availability  background  baddata  being  benchmarks  best  bias  big-peeps  bigdata  binary  biography  blocking  blog:  board  books  bostrom  brain-scan  brexit  bugs  calibration  cambridgeanalytica  capacity  capta  capture  causation  cct  census  cep  checking  checklist  checks  china  choice  circuit  cit  citations  cityjtips  classification  climate-change  cnc  coarse-fine  cog-psych  color  commodityquantity  comparch  comparison  competition  complex-systems  complexity  compsci  computation  computer-science  computer-vision  computers  computervision  concept  concurrency  conditions  confidence  confusion-matrix  consequence  consistency  conspiracytheory  context  control  cool  corrections  correctness  coverage  cpu  creditrating  crime  critique  cs  csun  cummingsdominic  dark-arts  dat  data-cleanliness  data  database  dataquality  dataset  datatypes  dataviz  datscan  death  deaths  debugging  decimal  decimals  decision  deconstruction  deep-learning  deep-materialism  deepgoog  deepvariant  deletion  dennett  density  design  desire  detail-architecture  developer  differrence  dikw  diplomatics  dirtydata  distortion  distributed  double  duplication  durability  duty  early-modern  echo  eden-heaven  efficiency  employment  ems  engineering  enhancement  enlightenment-renaissance-restoration-reformation  erasure  error  errors  essay  estimate  estimator  ethics  europe  evaluation  evidence  evolution  evopsych  expert  exposure  extension  face  facebook  facial  facialrecognition  fairness  fake  fascism  fcra  fiction  floating  floatingpoint  flux-stasis  for  forecasting  frontier  futurism  fuzzy  game  games  gb  gender  genetic...  genetic  geocoding  geodesy  geography  geolocation  geospatial  germanic  gettickcount  given  global  gnss  google  googleanalytics  government  gps  gpu  graphs  gsr  guidelines  gun  hacking  hanson  hardware  hatred  health  historicism  history  hmm  human-ml  human  ibm  ideas  idempotency  iee754  ieee-754  ieee  illusion  imaging  improve  improvements  in  incomplete  inference  info-foraging  information  innovation  interdisciplinary  interpretation  interval  intricacy  inventory  iqtrainwreck  issues  jacquesderrida  jargon  jmarkcooney  joannemcneil  jpl  justice  kit  knowledge  koganaleksandr  label  labour  lang:go  language  latency  learning  led  legibility  librarianship  library  lighting  lighting:  links  list  listening  listing  location-privacy  location  logistic  machiavelli  machine-learning  machine  machinelearning  machining  magnitude  map  mape  maps  marginal  martinheidegger  mask  matematiques  math  mathematics  meaning  measure  media  medicine  mentalhealth  meta:prediction  meta:reading  meteorology  methodology  metric  metrics  microsoft  mill  milling  misinterpretation  models  moderation  moderators  money  mortality  motivealgorithm  mouse  movies  music  mystic  nabilhassein  nasa  natural  netflix  neuro-nitgrit  neuro  neurons  news  nga  nibble  nitty-gritty  nlp  nonexpert  novelty  numbers  numerics  nytimes  object  objectivity  objektbuch  ocr  of  offense-defense  omission  open-problems  opencl  org:junk  overflow  overview  papers  paradox  parsimony  pd  pdf  performance  persistence  personaldata  petabyte  philosophy  photography  photos  php  pi  pic  pkgdown  plainlanguage  plots  pls  point  pointing  police  policing  policy  political-science  politics  polling  positioning  power  precise  precisio  precision  prediction  predictions  predictive-processing  predictive  preprint  presence  presentations  print  privacy  pro-rata  probability  processing  product  profiles  profiling  profjoekimble  program  programming  provenance  psychology  psychometrics  pyro  q-n-a  qualitycontrol  quiz  quorum  quotes  race  ranking  ratty  raw  rayleigh  reading  recall  recidivism  recognition  recommendations  reduction  reference  referendum  refusal  regulation  reinforcement  reliability  religion  report  representation  research  reviews  rg  rhetoric  rinex  risk  rms  roots  rounding  rudeness  said  sapiens  scale  scaling  sci-comp  science  screening  secrecy  selectcommittee  service  shiny  siggraph  simd  similarity  simulation  single  singularity  skunkworks  sla  sleep  socialmedia  software  sousrature  speculation  speed  speedometer  ssc  sse  sse2  stackex  state-of-art  statistics  stats  stock  strategy  stress  structure  studies  study  subscription  super  surveillance  surveillancecapitalism  systems  taken  targeting  technocracy  technology  technologyreview  telegraphy  test  the  theos  thinking  threat-modeling  time-series  time  tm112  to  tool  tools  top-n  trackerblock  trackers  tracking  tracks  trends  tribalism  trust  truth  tumblr  twitter  type:tool  types  typography  tyrieandrew  uber  uk  umbrae  uncertainty  under  unintended-consequences  us-them  useful  validation  verification  verify  video  visualization  volo-avolo  voteleave  warm  waveform  ways  weather  web  well  whitworth  wiki  wikipedia  windows  wire-guided  wisdon  within-without  work  wow  writing  written  your  youtube  yvain  |   

Copy this bookmark:



description:


tags: