Video Analytics Fundamentals Guide

By IPVM Team, Published Mar 04, 2021, 11:00am EST

Video analytics is a major differentiator in video surveillance, but because of the different marketing terms, names, and features, understanding how they work and differ is difficult.

IPVM Image

In this 30+ page guide, we teach the following:

  • Accuracy Concerns
  • Hardware Tradeoffs
  • Video Analytics And Computer Vision
  • Basic Image Analysis
  • Computer Vision Evolution in Video Surveillance
  • Video Motion Detection (VMD) - Pros and Cons
  • Heuristics Analytics - Pros and Cons
  • Machine Learning - Pros and Cons
  • Supervised and Unsupervised Learning
  • Haar and HOG
  • Deep Learning - Pros and Cons
  • Convolutional Neural Networks (CNNs)
  • Neural Network Optimizers
  • Don't Always Need Deep Learning
  • Datasets for Training
  • Common Datasets
  • Dataset Issues

This is the first chapter in our Video Analytics Course starting at the end of March.

Video ********* *******

***** ********* *********** ****** widely, **** ****** ** very **** *** ***** is ** ****** *** to ****** **** ***** on ***** ********* ******. Users *** *********** ****** be ******* ** **** analytics ***** ***** *** specific **********, ** **** how **** ** ******** performs. **** ***-********** ********* may **** ** *********** environments (*.*. ******* ********, bad *******).

************* **** ************ ********** performance *** ***** "*****" or "**" ********* **** struggles ** ****-***** ************. While ******-********** ********* *** available, **** *** **********, not *** ****.

Computational **** ** ******** *****-***

*** ******* ***** ********* accuracy *******, *** ******* the ************* ***** ***, generally. ***** ********** ***** (to ** ******* ** our *********** ********** ********), typically ********** ***** ********* accuracy ***** ********** ************* costs, ***** ** ****** of *********. *** *******, VMD *** **** **** more ******** **** * neural ******* *** ** might ******* */****** (** fewer) ********* *********.

*******************

******** *********** *** * major ****** ** ********** video ************ *********, ***** the ****-****, **** ****** nature ** *** ***********. Often, ******** ********** ***** analytics *** ******** ** fit *** ******** *********** of *******, ********, ***. See ****'****** ********* ******** ***** ***** ****.

Video ********* *** ******** ******

***** ********* ** * technology **** **** ********* to ******* ****** ** detect ** ******** *******. While **** ** * simple **** *** ******, computers ** *** "***" images ** ****** **.

**** * ******** ***** at * ********* ******* image, ** ** * collection ** ******:

IPVM Image

* ******** *** *** advantage ** ***** **** to ****** ******* ** many ****** **************, ********* humans *** ******* ******* in. *******, ********* **** a ************ ** **** it ***** *********** ********* to ** *** **** of ********* **** ****** take *** ******* ** even ** * *******.

***** ********* ***** ********** or ******* ** ****** and ***** ** **** patterns ** ******* ** the ****** **** ********* an ****** (*.*. ******, Face, *******), ******** (*.*. a ****, * ****, a *****), ***/** ********* (e.g. *********, ******* ********, fighting).

Basic ***** ********

***** ********* ******* *** grayscale ****** ** **** differences ** *** ******, which ********* ***** ** objects:

IPVM Image

********* ****** *** *** smaller **** ***/***** ******, with *** **** ****** of ******. ******* ****** allow *** ***** ********** resources, ********** **** *****.

***** *********** ** *** values ** * ******** pixels *** ******** ** edge, *** ******** ****** with ******* ****** ** not:

IPVM Image

***** ****** ********** *** at *** **** ** all ***** ********* *** are ******* ****** **** in *** **** ***** ones **** ** ********* VMD.

Computer ****** ********* ** ***** ************

***** *** * **** analytic ********** *** ***** analytics, ** ***** ** complexity:

  1. ***** ****** ********* (***)- ******* ******* ** pixels, *** *******
  2. **********- ******* *** ****** based ** ****-***** *****-******* rules (*.*. "****** ***** changes ******* *", "****** pixel ******* **** **** last * *******")
  3. ************ ****** *********- ******* ******* ***** hybrid *****-******* ***** *** basic ******* ******** (******** optimizes *** ********* ***** on ******** ******).
  4. **** ******** ****** *********- ******* ******* ***** computer-generated **********, ***** ** training ****/******.

Video ****** ********* (***)

*** ******** ******* ** the **** ***** **** one *****/***** ** *** next. ** ****** ******' color/contrast ****** ** ****** of * ***** (*.*., sensitivity), *** ****** ******** motion ********. *** **** not ****** *******, ****** the ***** ***** ********* covered ** **** ******, and **** *** ******* values ** ******** ******.

*** ***** ***** ***** how *** ***** '***' a *** ****** *** correctly ********* ******:

VMD **********

*** *** * *** computational *** *********** ****, essentially ****, *** ** ubiquitous ** ** *******.

****, ** **** ***** scenes, **** ******* ***** changes, *** *** ********* detect ** ***** ** no ******, ******* ** people ** ******** *** detected.

IPVM Image

*******, *** *** *********** fundamental ******.

VMD *************

*** ** **** ***** to ******, *** ******** will ******* ******* ***** on ***** ******* *** to ********/*******, ******* *** cannot ****** ** ****** is ***** ** ******:

*** **** ********* ** outdoor ************, ********** ********** due ** *******, *******, and *** *********** (*.*. trees, ****** ** *****):

IPVM Image

Heuristics *********

********* ******* *** ***** to *** ** ******** false ****** ***** **** keeping *********** *** **** low. ******** *********, ********* camera ************* ***** * series ** ******* ** options ** **** ********* some ***** ******.

*** **** ****** ******* are ***** ****** ***********, duration ** ********, *** direction ** ********, *** a ******* *** ******* size:

IPVM Image

* **** ******** ***** of ********* ********* ** factoring ****** ***** ********* or ****** ** ***** ratio ** ********* **** type ** ****** ** in ****** (*.*. ****** or *******):

IPVM Image

Heuristic ****

********* ********* ** ******* to *** ** **** they *** ********** **** to *******, *** *** n** *************** ****, *** are **** ******** **** VMD:

IPVM Image

********, ** ********** ****** scenes, ********* ********* *** accurately ****** ** ***** is ** ****** ******* and **** ** **** prone ** ***** ****** due ** ***** ***** changes **** ***.

Heuristic ****

********* *********' ******* ******** is **** **** *** limited ** ********* ******** or ********* **** *** hard-coded ** ******.

********* ********* *** ***** to ****************** **** ******* do *** **** *** pre-set ************, * ****** crawling ** *** ****** wearing ****** ******* ******** might ** ********** ** a ******* ******* ** a ******, ***** ***** aspect ***** *** ******* coloring:

IPVM Image

************, ********* ******* ******* more *****, *****-***-***** *******, to ******** ***** ******** than **** ******** *********.

Machine ********

******* ******** ** *** next **** ****** ********** in ***** *********. ***** a ***** ******** ***-********** instructions ******* **** *** computer ** **** **, the ******** ********* *** algorithm, *********** ***** **** more ********* **** ***** be ********** **** ********* models. ** ****, ******* learning ********* *** ********* more *************** ******* *** more ********.

*** *******, * ******** is **** *** ** detect * ****** ******* by *********** *********** **********:

  • *** ***** ** *** height ****** ** ****** than *** *****
  • ***** ****** ** ******** of **** *** **** that ** **********
  • ***** ****** ** *********** movement ******* ** ********
  • ***** ****** ** **** color *** ******* ******* (representing ********)

**** *** ********* ** then *** *****, ** will **** *** ***** parameters, *** ** ** finds ****** ** ****, it **** ****** *** video ******** * ****** walking.

Supervised *** ************ ********

**** ***** ************ ********* use ********** ******** ****** development. ********** ***** *** training ****** *** *******, and *** ******** ******* what *******/****** **** ** used ** ****** *** objects. ******** ******* ****** are ******** *** ******* learning ** ****** *** correct *******.

********* ** *** ********, one ** **** ****** can ** **** ** provide ******* ****** **** meaning *** *********** ** the ******. **** ***** pictures ** **** *** be ******* ** ****, and ************ ** ******* as *** **** ** gun:

IPVM Image

** ************ ********, *** images *** *** *******, and *** ******** ******* how ** ***** *** objects. *******, **** ***** the ******* **** ***** in ** ************* ******, with *** ****** ******* of ********* ******** ********, like *** ***** ** objects ******* ** ****** or *******:

IPVM Image

************ ********, ** ********, would ******* **** **** by *********** ******** ******** of ****** ** ********. Additionally, ************ ******** *** find **** ******* ******** and ******** **** * human ***** *** ********* or ******. ** ****, unsupervised ******** ** ********* applicable ** ********* ********* in ****-**** ******** ** movement-related ********.

***** ****, ************ ******** use ** ******* ** video ************ *** ********** or ******* ******** *********. See ****'* ***** ********** ******* ******** ******************* ********* ********* ******* Detection.

Haar *** ***

**** *** ***-***** ******* learning ********* *** ******* in **** **** *** human-defined ******* *** *** opposite ********** ** ****** objects. **** ***** ********, then *****. *** ***** edges *** **** ********.

****

**** ** *** ** the ****** ******* ******** methods *** ****** *********, is *************** *********, *** commonly **** ** ** cameras.**** ** **** **** at ********* ***** *** lines ** ******* *** uses ********* ******* *********** (eyes ********* ** * nose, ******** ***** *** eyes) ** ****** ***** and *** ******* ** the ****:

IPVM Image

** *** ***** *******, this ****** *** ********* eyebrows ***** *** ***:

IPVM Image

**** *********** *** ******* windows *** ******** ******* filters, ** **** ** 38 ** *** ** the **** ****** ******* Viola-Jones.

IPVM Image

******* **** ***** ** finding *****, ** ******** well-lit ******** **** * high ***** ** ******, and**** **** ********* ************ ** ********** ********* with ****** *****.

***

*** ** ******* ****** machine ******** ********* ****** used ** ***** ************, offering * ******** ********** between ********** *** ********. HOG ****** *** "********* of ******** *********" *** detects ******* ** ******* edges *** ******* ** an *****, *** *** strong/sharp *** ***** *** corners ***:

IPVM Image

******* *********** *** ******* to *** ********** ***** and *** ** ********** to ****** *** ***, persons, ** ******** ******* like *** ******* *****, by ***** *** ******** to **** *****. ******* of ****, *** *** be **** ******** **** Haar ** **** *****, and **** ***** ** errors *** ** ******** and ****** ******** ** Haar.

*** **** ******* ** face ********* ***** ****, HOG, *** **** ********, ee **** ****'***** ********* ********.

Machine ******** ****

******* ******** ********* *** computer-optimized *** **** **** features **** *****-********** **********. This ***** ** ** more ******** ** ********* objects **** **** ** not ***** ***** *********** or ********** ********. ** such, ******* ******** ********* are ****** **** ********** at ********* * ****** whether **** *** ********, running, *******, ***..

******* ******** ******** **** processing **** *** *** heuristics *** ** ********* efficient ****** ** ** supported ** ***** ********** powered ******* **** ******* and ****. ********, **** have ***** ************ **** deep ********, *** **** are *** ****** *** most ************.

Machine ******** ****

******* ******** ** **** costly ** ******* ******* the ******** ***** ***** amounts ** ****/****** ** learn ****, ***** ** expensive. ************, ******* ******* learning ** **** ******* than *** ** ********* analytics, ** ** **** computationally ********* *** ** camera-based ******* ********* **** (see ********* ********* ******** ***).

******* ******** ********* *** also **** ******** **** accuracy **** ******* ** a ****** ******* ********** change, *** ******* **** a **** ** ******* with * ****:

IPVM Image

** ******** ******** *** analytics ************ ****** ******* Learning, ******** ********* ******* Deep ******** ********** ** video ************.

Deep ******** ********* ***** ****** ********

**** ******** ********* ******* images ******* *+ ****** of ******* ** ************** neurons, ***** ********** **** the ******** ****** **** likely **:

IPVM Image

***** **** ******** ********* learn ** ***** *** what ** ***** ** during *** ******** *******, the ***** ** *** network ** ******* ** the *********, ********* ******* trial-and-error.

*******, ****** ********** ** machine ********, *** ******** rules ** ******* *** not ********** ** ******, they *** ********** ** the ********, ***** ** the ******** ******.

Learning **** *** ****** ** *******

* ****** **************** ** that **** ********/** ********* continue ** "*****" ***** installation. **** ** ******** in **** *****, **** as ********'* ****-******** *** cameras, *** **** ******** analytics, ***** ******* ** learn *** ***** **** are ****** **.

*******, *** **** *********, all ******** ******* ****** the ******** ** ********, and **** *** *******-******* model, ***** **** *** change **** **** ***** on ******** ** ******* detected ** *** ***** of ****. ************* *** periodically ******* ***** ****** requiring * ******** ******* on **** ****** ** apply *** ******* *****.

Convolutional ****** ******** (****)

************* ****** ******** ** CNNs *** **** ******** used *** ******** **** learning ***** ********* *** object ********* *** **************.

*** * ******** ****** of * **** ***:

  • ***********:******** ********/******* **** ** image ** ******** * sliding ****** ******, ******* to **** ******* **********, which ******* *** ************* cost ** *** *********.
  • ***-********* / ****: ***-****** ********** ** the *********** ****** ** make **** *** ****** network *** ******* *** non-linear ******** (*.*. ***** of ****** *** ***** of ********* ********).
  • *******:************* ******* *** ***** image **** ******* **** convolution ***** ** *********** multiple ******* **** *** representative ******. **** ****** the ************* **** ** reducing *** ****** ** information ****. ** **** helps **** ********** ** small ******* ** ********, angle, ** ********** *** creates * ****** ******** to *** **** ** overfitting, ** ************ *** information ** ********* ***/******** reducing *** ***** ** detail.
  • ***** *********: **** *** ******** from *** *********** *** Pooling ****** ** ******** the ***** **** ***-******* objects/outputs, ***** ** *** training ********.

IPVM Image

** **** **** ** finer ******* ** ** image, *********** *** **** layers/filters (******* ** *** vs *** *****). *******, due ** *** ******** ways **** ******** *** be ***********, *** ***** of *** ****** ******* is *** ****** * good ********* ** ******* system ********. ********, **** layers ** * ****** network **** ********* ******* more ******** ********** *** higher ************* **** *** camera:

IPVM Image

Open ****** **** - ******** *** ****

******** **** ** ****-****** CNN *** ***********/********* ** for * ******** *********** is ****** ** ***** analytics. ************* ***** ****** and ********* ****-****** ****, saving *********** *** ************.

**** ***** ************ ************* may *** **** *** data ******* ** *********** resources ******** ** ***** their *** ****** ********, and **** ****-****** **** (e.g.*********,******,*********,****, ***) *** ****** accurate *** *************** *********.

Neural ******* **********

********** *** ********** ** models ***** *********** **** accuracy ** *** ** the **** ******** ***** for **** ******** *********. Neural ******* ********** *** algorithms ** ******* **** analytic ********** *** *** to ****** ** ****** how *** ******** ****** during *** ******** *******.

***** *** **** ************/*********** methods *** ********** ****** network ******** (*.*. ******** Descent, **********, ******** ********, etc.) **** ********* ************* costs *** ******** *****-****.

*****'* ********** * ********-******** ********* that ************* ********* *** efficiency ** ******* **** learning ***** ********* ** Intel ***/****, ** ********** the ****** ** ***** the ****** *******:

IPVM Image

Deep ******** ****

**** ******** ********* ********* offer ************* **** ******** performance *** *** *** susceptible ** ***** ****** due ** **** ******** or ***** *******. ************, they *** ****** **** to ******** ******* ********** in *********** **********, ** when *** ****** **** not ***** *** ***** appearances ** ************:

IPVM Image

**** ************ ********* ************ are ***** ** ****** objects ** * **** or ***** ***** **** should *** ** (*.*.: trespassing ********* *****-*****). **** learning ********* *** **** to ****** "********" ********, people ******* **** **** usually ****, ** * large ***** *** ****** in *** **** ********* suddenly (**** ** ** a ***** **********), ***** other ***** ** ********* would ** ****** *******. By ***** **** ** better ********* ******* ** interest ** * *****, deep ******** ***** ***** analytics ******** ***** ******** and **********.

Do *** ****** **** **** ********

*** **** *****, **** people ** **** ********* on ** ****** ******* intercom, * ******* ******* learning ********* ** **** enough ******* *** ***** of **** ** ******, the ****** ** ******* directly ** *** ******, and ******** *** ** controlled.

IPVM Image

***** ******* ********, *** even **** ** ** application, **** **** *********** costs **** ******** ** deep ********, *** **** provide ******* **** ******** detection ** **** *********** environments.

Deep ******** ****

***** **** ******** ********* are **** ******** **** heuristic ** ******* ********, they ********* ******* ********* GPU ******** (*** ********* ********* ******** ***) ******* ** ************* requirements *** ********, *** in **********.

************, ******* **** ******** analytics ******** * *** of ******** ****/******, ***** development *** ******** ***** can ************* ******** *** final **** ** *** analytics ** *****. **** learning **** *********** *** cost ******* ** $*,*** - $*,*** *** ******.

********, ***** **** ******** analytics *** ********* **** at ********* *** *********** objects **** ** *** exactly ***** ******** ****** or ********** ********** *******, every *********** *** ********* challenges. ** ****, ***** should ** ********* **** multiple ***** ** *** real-world ********, ** ******* performance ** * **** range ** *******, ********, and *******.

Datasets *** ********

******** *** **** ** train **** ******** ********* on *** ** ****** or ********* *******, ********, behaviors, *****, ** *** object/action. ******** *** ********* composed ** ********* ** millions ** ******* ****** or ******.

**** ******** ********* ******** is ******* ********* ** dataset ******* ***** ** 4 *******:

  • ******* ******
  • ************** ** *** ****
  • ******* ******
  • ****** ** ******

******* ******

******** ******* ******** ****** or * ******** ******* (e.g. ****** ***********, ******* detection):

IPVM Image

*******, **** **** **** to ** ********* *******, and ******* ** ** so *** ****** ** training ** ********* ** detect ** ********* *** wrong *******/******** (*.*. **** length ******* ** ******, snowy ******* ******* ** dogs ******* *******).

************** ** *** ****

***** **** ******** *** created **** ***-************ ****** (e.g. ***** ******, ********, passport/ID ******), ** ** important ** ***** ************ analytics **** ************ *****. Surveillance ******* ********* **** challenging ****** *** ********.

***** **** ******** ********* is **** ********* ** these ****** **** ******* learning, ******** **** ****** decrease *** **** ***** if ************** **** ** not ****:

IPVM Image

*******, ************-******* ******** *** not ******, ******* ******** may **** ** ** created ** *** ******** developer, ** * *********** cost.

******* ******

******** **** ** **** images **** *** ******* with ******* ** ******** like *********, ***, *** gender, *** ******* ** do ** **** ** bias, ***** ********, *** risk ** ****** ********. Additionally, ******* ******* ****** can ** ********* ** get, **** ****-****** ********* of ***** ************* **** in *********:

IPVM Image

********, *******, ********, ** clothing *** **** **** different ** *** ** vs ***** ** ** South *******. ********, * system ***** ** ******* with ********* ***** ** European, ** ******* *****. Such * ****** ******** in ******* **** ****** of ********* ********, ******, etc. *** **** ** perform **********.

*** *****

******** **** ******** **** are *** ***** **** create **** ******** **********. Algorithms **** ** ***** that * ****** ** the **** ****** ******* aging, ********** ******, ******** change, ****-***** *******, ********, angle, *** ****-** ***** differentiating ******* "*************" ***** requires * ***** ****** of ****.

Common ********

***** ****** ******* ***** used *** ********* ********, ranging **** **** ******** available ******** ** ******* government *** ************-******* ********.

****** ****** ********

****** *** ** *** more ******** **** ****** datasets, ******* ** ** large, **** ********, *** includes **** ****** *******. It ******** ***,*** ******, with *.* ******* ******* objects ** ** **********, which *** ** ******** for ******** ************:

IPVM Image

********** ******* ****** **** public ******* *** ************** training, **** **** ** million ******, *** **** 100,000 **** *********** ****** "synset" ** ***** ******* in ******. ******* *** typically ******* **** ******** synsets ** ******** ********** levels ** ******

IPVM Image

********* * ****** ******* that ** ***** *** simpler **** **** ** Imagenet. *******, ** ** good *** ******** ****** detection, *** ******** ~**,*** labeled ****** *** *** labeled ****** *****:

IPVM Image

******* * ****** ******* focused ** **** *********, containing **,***+ ****** *** 390,000+ ******* *****. ***** labels "**** **********" ********* poses, **********, *** ************:

IPVM Image

********** ********

*********** **** ****** ** high-quality ****** ** ***** population, ******* ************** ****** (e.g. ******* *******, ********, etc.) *** *******/******* ******. However, ********** *** ** these ****** ******. *** PRC ****** ********** *** openly ** ****** *********** companies *** *********** *** testing. *** ** ********** uses ******** *** ******* IDs *** ******** ******* (e.g. ****).

*******, ** *** **,******* ***** ******* ********** varies **** *****-**-*****, *** *** ********* public ****** **** * person ** ********* ** a *****. **** ********* offer ****** ********* ** US ******** (*,*,*), ***** ***** ** accessed *** ****** *********** training.

*********** ********

*** ** *********** ************ (e.g. *** *********, ******** detection), **** ******** ********* tout ***** *** ** proprietary ********. ***** ***** might ** ****** ********, offering ****** ******** ** training, ***** ** ** way ** ****** *** diversity ** ******, **** of *** ********, ** ethical/legal **********.

Dataset ******

***** *** ****** ********** and ********* ********** ** video ************ **** **** to ** ********* **** building ********. ****** **** video ********* ******** *** caused ****** *** ******** process ******* ** ********* needs ********** *** ******** data ** ******** ***** what ** ****** *** classify.

Not ****** ******

******* ****** ******, *** algorithm *** ***** ** details **** *** *** important ** *** ***********, and ****** *******, *** useless *******. *** *******, a ****** *********** ****** may ****** ****** ******* similar ********, ****** **** classifying ******* **** ********:

IPVM Image

Distinct ******** *** ** ******* ******

******** ******** ** * person ** ******* *** be ****** *******, *** an ********* **** ****** heavily ** ****, ****** hair, *******, ********, ***. may ******* ****** ** the **** *****:

IPVM Image

***** ******* ****** ******** (e.g. *******, ******), ** not ***** * *********** decrease ** ********, **** testing *** ******** ******** ************* ***** **** length ********** ****** **************:

IPVM Image

Public ******** - ***** *** ******* **********

****** ******** **** ***** challenges, ********* ** *** US *** ****** *** to ******* ******** *** government **********. *** ***** challenges **** ****** *** use *** *********** ** facial *********** **********, ** limiting *** ********** ** faces **** *** ** used *** ******** ****, and ******** ** ****** recognition *******.

****** *********** ********** *** also ****** ******* **********. While ***** ****** ***** and ********** ** ****** may ** *****, ** is *** ***** ** many ****** **** ***** faces *** ** **** for ****** *********** ********. In ****, ******, *********, and *********** **** ****** *********** development********* ****** ********.

Labeling **** ** *** ****

******** ****** ** ******** takes * *** ** time ** ** **********, because ** *** ******** of ********* ** ******** of ****** ****** ** datasets.

IPVM Image

************, ***** *** ********** and ******* ********** ** labeling:

  • ** ** ******* ** add *********, ******, ********* labels, ***.?
  • * ******'* **** *** no *** **********, *** it ** *******?
    • **** ***** ** **** for **********?
  • **** ******* * **** of *******?
    • * *******? ** ******** bicycle? * ********?

*** ****** *** ******* datasets *** ******** ************ has ****** ********* ** offer ******* ******** ******** through ********* ********** ********** *********** ******** ** *** highly-skilled, *** ********* *****-*********.

Comments (18)

**** *******. **** *******.

Agree: 6
Disagree: 1
Informative: 1
Unhelpful
Funny

IPVM Image

*** *** ****** ****, or ****, ****** ** be ********** ** * human? ****, **** ****** detection ****** ***** ***** pretty ****.

Agree
Disagree
Informative: 1
Unhelpful
Funny: 6

**** ***** ** *** graphic *****/****.

Agree
Disagree
Informative: 1
Unhelpful
Funny

***** ** ******!

Agree: 1
Disagree
Informative
Unhelpful
Funny
  • *******:************* ******* *** ***** image **** ******* **** convolution *****. **** ***** to ******* *** ****** network **** "***********" ** only ***** **** ** detect ******** ******** ******, instead ** *** ******, vehicle, ****, ***.

* ******* "*******" **** down ******* ******* **** between ****** ** ******* overall **********. "***********" ****** result **** *** *** architecture *** ******** ****/******* used, *** *******.

Agree: 1
Disagree
Informative
Unhelpful
Funny

**** ********. **********.

Agree
Disagree
Informative
Unhelpful
Funny

**** ***. ******

Agree
Disagree
Informative
Unhelpful
Funny

***** ***** - ******!

Agree
Disagree
Informative
Unhelpful
Funny

****: "** ********** *** number ** ********** ****** *******:"

Agree
Disagree
Informative
Unhelpful
Funny

**** ***, ******.

***** ** * **** in **** ******:

**** ******** ********* ******** is ******* ********* ** dataset ******* ***** ** 3 *******:

*** ************, ****.

Agree
Disagree
Informative
Unhelpful
Funny

***** *** *** ******** that *** *** * am **** *** ******* the ****** *****. ** have ***** **.

Agree
Disagree
Informative
Unhelpful
Funny

* *** ** **** info ****** *** ********* the ********.

Agree
Disagree
Informative
Unhelpful
Funny

***** *** ** *** much *** ****** ** such ******** ****** ** video ********* ************. ** a ******** ** *** very **** ** ********** and **** ******** ** well. **** ** **** information ** ********* *** mentioned ** ** * little ******** ** *** first ****** (* ******* on **********).

*******************, ** **********(/**ʊəˈ*ɪ**ɪ*/;******* *****:εὑρίσκω,****í**ō, '* ****, ********'),** *** ******** ********* *************-************* ******* * ********* method **** ** *** guaranteed ** *********, *******, **********, *** ** ************ sufficient *** ******** ** immediate, *****-**** **** ***************. ***** ******* ** optimal ******** ** ********** or ***********, ********* ******* can ** **** ** speed ** *** ******* of ******* * ************ solution. ********** *** ** mental ********* **** **** the********* ************ * ********. (******* ** *********)

***** ********* **** **** a ***** **** ** research ***** *** **** few ******* **** ******* techniques **** ********* ** overcome ********** **** ** accuracy *** ********* ** analyzing ***** ****. **** section ******* *** **** concept ** **** ***** works *** **** ****** works ******** ** **** learning. *** **** ************ foundations ********* ** **** section *** ** ******* to ********* *********** ****** such ** ****** ********, segmentation *** ****** ********* to **** * ***.

Here ** * ****** *********** *** ********** (* ****** ****) ******* ***** ********* ****** *******:

** *******, ********* ********** is ********* *** * single ********** ** ** because *** ****** **** never ** **** ** find *** *** ********* problems ** ** *********. Luckily, ********** **** **** different ******** *** ***** that ********* ****** **** different ********* ********. *********, it ** ******** ** improve *** ************* ** the ****** ************* ** involving ******** **********. ****** 1 ***** ** ******* from * **** ***** of ********* ********** ***** 19 ********** **** **** to **** ** ********* problems ** * ***** response ****** ******** ********* access ** ***** **** accounts (******* ****). **** of *** ***** ******* in ****** * ********* the ******* ** *** of *** ********* ******** by *** ** *** evaluators. *** ****** ******* shows **** ***** ** a *********** ****** ** nonoverlap ******* *** **** of ********* ******** ***** by ********* **********. ** is ********* **** **** some ********* ******** *** so **** ** **** that **** *** ***** by ****** *********, *** there *** **** **** problems **** *** ***** by **** *** **********. Furthermore, *** ****** **** identify *** **** ********* and **** ****** ** that ******'* ********. *****, it ** *** *********** true **** *** **** person **** ** *** best ********* ***** ****. Second, **** ** *** hardest-to-find ********* ******** (*********** by *** ******** ******* in ****** *) *** found ** ********** *** do *** ********* **** many ********* ********. *********, it ** ********* ** involve ******** ********** ** any ********* ********** (*** below *** * ********** of *** **** ****** of **********). ** ************** is ******** ** *** three ** **** ********** since *** **** *** gain **** **** ********** information ** ***** ****** numbers.

Figure * ************ ******* ***** ********** found ***** ********* ******** in * ********* ********** of * ******* ******. Each *** ********** *** of *** ** ********** and **** ****** ********** one ** *** ** usability ********. **** ****** shows ******* *** ********* represented ** *** *** found *** ********* ******* represented ** *** ******: The ****** ** ***** if **** ** *** case *** ***** ** the ********* *** *** find *** *******. *** rows **** **** ****** in **** * *** that *** **** ********** evaluators *** ** *** bottom *** *** ***** successful *** ** *** top. *** ******* **** been ****** ** **** a *** **** *** usability ******** **** *** the ******* ** **** are ** *** ***** and *** ********* ******** that *** *** **** difficult ** **** *** to *** ****.
IPVM Image

********* ********** ** ********* by ****** **** ********** evaluator ******* *** ********* alone. **** ***** *** evaluations **** **** ********* are *** ********** ******* to *********** *** **** their ******** **********. **** procedure ** ********* ** order ** ****** *********** and ******** *********** **** each *********. *** ******* of *** ********** *** be ******** ****** ** written ******* **** **** evaluator ** ** ****** the ********** ********* ***** comments ** ** ******** as **** ** ******* the *********. ******* ******* have *** ********* ** presenting * ****** ****** of *** **********, *** require ** ********** ****** by *** ********** *** the **** ** ** read *** ********** ** an ********** *******. ***** an ******** **** ** the ******** ** **** evaluation *******, *** ******* the ******** ** *** evaluators. ****, *** ******* of *** ********** *** available ****** **** ***** the **** ********** ******* since *** ******** **** needs ** ********** *** organize *** *** ** personal *****, *** * set ** ******* ******* by ******. ***********, *** observer *** ****** *** evaluators ** ********* *** interface ** **** ** problems, **** ** ** unstable *********, *** **** if *** ********** **** limited ****** ********* *** need ** **** ******* aspects ** *** ********* explained.

** * **** **** situation, *** ******** (******** called *** "************") *** the ************** ** ************ the ****'* ******* ** order ** ***** *** these ******* *** ******* to *** ********* ****** in *** ****** ** the *********. **** ***** it ******** ** ******* user ******* **** ** the ***** ** *** know ******** ***** **** interface ******. ** ********, the ************** *** ********* the **** ********* ** placed **** *** ********* in * ********* ********** session, ** * ******** observer **** ***** ** record *** *********'* ******** about *** *********, *** does *** **** ** interpret *** *********'* *******.

*** ******* *********** ******* heuristic ********** ******** *** traditional **** ******* *** the *********** ** *** observer ** ****** ********* from *** ********** ****** the ******* *** *** extent ** ***** *** evaluators *** ** ******** with ***** ** ***** the *********. *** *********** user *******, *** ******** wants ** ******** *** mistakes ***** **** **** using *** *********; *** experimenters *** ********* ********* to ******* **** **** than ********** *********. ****, users *** ********* ** discover *** ******* ** their ********* ** ***** the ****** ****** **** by ****** **** ******** by *** ************. *** the ********* ********** ** a ******-******** ***********, ** would ** ************ ** refuse ** ****** *** evaluators' ********* ***** *** domain, ********** ** ********* experts *** ******* ** the **********. ** *** contrary, ********* *** **********' questions **** ****** **** to ****** ****** *** usability ** *** **** interface **** ******* ** the *************** ** *** domain. *********, **** ********** have ******** ***** *** interface, **** *** ** given ***** ** *** to ******* ** ***** not ** ***** ******** evaluation **** ********** **** the ********* ** *** interface. ** ** ********* to ****, *******, **** the ********** ****** *** be ***** **** ***** they *** ******* ** trouble *** **** ********* on *** ********* ******* in ********.

*********, * ********* ********** session *** ** ********** evaluator ***** *** ** two *****. ****** ********** sessions ***** ** ********* for ****** ** **** complicated ********** **** * substantial ****** ** ******** elements, *** ** ***** be ****** ** ***** up *** ********** **** several ******* ********, **** concentrating ** * **** of *** *********.

****** *** ********** *******, the ********* **** ******* the ********* ******* ***** and ******** *** ******* dialogue ******** *** ******** them **** ***** ** ********** ********* principles(*** **********). ***** ********** are ******* ***** **** seem ** ******** ****** properties ** ****** **********. In ******** ** *** checklist ** ******* ********** to ** ********** *** all ******** ********, *** evaluator ********* ** **** allowed ** ******** *** additional ********* ********** ** results **** **** ** mind **** *** ** relevant *** *** ******** dialogue *******. ***********, ** is ******** ** ******* category-specific ********** **** ***** to * ******** ***** of ******** ** * supplement ** *** ******* heuristics. *** *** ** building * ************* **** of ********-******** ********** ** to ******* *********** ******** and **** ******* ** existing ******** ** *** given ******** *** *** to ******** ********** ** explain *** ********* ******** that *** ***** (******* 1993).

** *********, *** ********** decide ** ***** *** how **** **** ** proceed **** ********** *** interface. * ******* ************** would ** **** **** go ******* *** ********* at ***** *****, *******. The ***** **** ***** be ******** ** *** a **** *** *** flow ** *** *********** and *** ******* ***** of *** ******. *** second **** **** ****** the ********* ** ***** on ******** ********* ******** while ******* *** **** fit **** *** ****** whole.

***** *** ********** *** not******** ****** ** **** (to ******* * **** task), ** ** ******** to ******* ********* ********** of **** ********** **** exist ** ***** **** and **** *** *** been *********** (******* ****). This ***** ********* ********** suited *** *** ***** in *** ********* *********** lifecycle.

** *** ****** ** intended ** * ****-**-***-*** interface *** *** ******* population ** ** *** evaluators *** ****** *******, it **** ** ******** to *** *** ********** use *** ****** ******* further **********. ** *** system ** ******-********* *** the ********** *** ****** naive **** ******* ** the ****** ** *** system, ** **** ** necessary ** ****** *** evaluators ** ****** **** to *** *** *********. One ******** **** *** been ******* ************ ** to ****** *** ********** with * ******* *************, ******* *** ******* steps * **** ***** take ** ******* * sample *** ** ********* tasks. **** * ******** should ** *********** ** the ***** ** * task ******** ** *** actual ***** *** ***** work ** ***** ** be ** ************** ** possible ** *** ******** use ** *** ******.

*** ****** **** ***** the ********* ********** ****** is * **** ** usability ******** ** *** interface **** ********** ** those ********* ********** **** were ******** ** *** design ** **** **** in *** ******* ** the *********. ** ** not ********** *** ********** to ****** *** **** they ** *** **** something; **** ****** ******* why **** ** *** like ** **** ********* to*** ************ ** ***** ********* results. *** ********** ****** try ** ** ** specific ** ******** *** should **** **** ********* problem **********. *** *******, if ***** *** ***** things ***** **** * certain ******** *******, *** three ****** ** ****** with ********* ** *** various ********* ********** **** explain *** **** ********** aspect ** *** ********* element ** * ********* problem. ***** *** *** main ******* ** **** each ******* **********: *****, there ** * **** of ********* **** *********** aspect ** * ******** element, **** ** ** were ** ** ********** replaced **** * *** design, ****** *** ** aware ** *** *** problems. ******, ** *** not ** ******** ** fix *** ********* ******** in ** ********* ******* or ** ******* ** with * *** ******, but ** ***** ***** be ******** ** *** some ** *** ******** if **** *** *** known.

********* ********** **** *** provide * ********** *** to ******** ***** ** the ********* ******** ** a *** ** ****** the ******** ******* ** any *********. *******, ******* heuristic ********** **** ** explaining **** ******** ********* problem **** ********* ** established ********* **********, ** will ***** ** ****** easy ** ******** * revised ****** ********* ** the ********** ******** ** the ******** ********* *** good *********** *******. ****, many ********* ******** **** fairly ******* ***** ** soon ** **** **** been **********.

*** *******, ** *** problem ** **** *** user ****** **** *********** from *** ****** ** another, **** *** ******** is ********* ** ******* such * **** *******. Similarly, ** *** ******* is *** *** ** inconsistent ********** ** *** form ** *****/***** **** formats *** *****, *** solution ** ********* ** pick * ****** ************* format *** *** ****** interface. **** *** ***** simple ********, *******, *** designer *** ** *********** to **** ****** *** exact ******* ** *** interface (*.*., *** ** enable *** **** ** make *** ****** ** on ***** ** *** two **** ******* ** standardize).

*** *********** *** ********* the ********* ********** ****** to ******* **** ****** advice ** ** ******* a ********** ******* ***** the **** ********** *******. The ************ ** *** debriefing ****** ******* *** evaluators, *** ******** **** during *** ********** ********, and *************** ** *** design ****. *** ********** session ***** ** ********* primarily ** * ************* mode *** ***** ***** on *********** ** ******** redesigns ** ******* *** major ********* ******** *** general *********** ******* ** the ******. * ********** is **** * **** opportunity *** ********** *** positive ******* ** *** design, ***** ********* ********** does *** ********* ******* this ********* *****.

********* ********** ** ********** intended ** *"******** ********* ***********"******. *********** ******** (******** et **. ****) *** indeed ********* **** ********* evaluation ** * **** efficient ********* *********** ******. One ** ** **** studies ***** * *******-**** ratio *** * ********* evaluation ******* ** **: The **** ** ***** the ****** *** ***** $10,500 *** *** ******** benefits **** ***** $***,*** (Nielsen ****). ** * discount ********* *********** ******, heuristic ********** ** *** guaranteed ** ******* "*******" results ** ** **** every **** ********* ******* in ** *********.

*********** *** ****** ** Evaluators

** *********, ********** ********** can ******* * ********* evaluation ** * **** interface ** ***** ***, but *** ********** **** several ******** ********* **** fairly **** ******* *** achieved **** ******* ** single **********. ******** **** six ** ** ********, single ********** ***** **** 35 ******* ** *** usability ******** ** *** interfaces. *******, ***** ********* evaluators **** ** **** different ********, ** ** possible ** ******* ************* better *********** ** *********** the *********** **** ******* evaluators. ****** * ***** the ********** ** ********* problems ***** ** **** and **** ********** *** added. *** ****** ******* shows **** ***** ** a **** ****** **** using **** **** *** evaluator. ** ***** **** reasonable ** ********* *** use ** ***** **** evaluators, *** ********* ** least *****. *** ***** number ** ********** ** use ***** ****** ** a ****-******* ********. **** evaluators ****** ********* ** used ** ***** ***** usability ** ******** ** when ***** ******* *** be ******** *** ** extensive ** *******-******** *** of * ******.

Figure * ***** ******* *** ********** of ********* ******** ** an ********* ***** ** heuristic ********** ***** ******* numbers ** **********. *** curve ********** *** ******* of *** **** ******* of ********* **********.
IPVM Image

******* *** ******** (****) present **** * ***** based ** *** ********* prediction ******* *** *** number ** ********* ******** found ** * ********* evaluation:

*************(*) = *(* - (1-l)*)

******************(*)********* *** ****** ** different ********* ******** ***** by *********** ******* **************** **********,********** *** ***** ****** of ********* ******** ** the *********, *** * indicates *** ********** ** all ********* ******** ***** by * ****** *********. In *** **** ******* (Nielsen *** ******** ****), the ****** ********* **** ** ******* to ** ******* **** a **** ** ** percent. *** ****** ********* **** ** ** 50 **** * **** of **. ***** **** formula ******* ** ****** very **** **** **** shown ** ****** *, though *** ***** ***** of *** ***** **** vary **** *** ****** of *** ***************, ***** ***** **** vary **** *** *************** of *** *******.

** ***** ** ********* the ******* ****** ** evaluators, *** ***** * cost-benefit ***** ** ********* evaluation. *** ***** ******* in **** * ***** is ** ********** *** the **** ** ***** the ******, *********** **** fixed *** ******** *****. Fixed ***** *** ***** that **** ** ** paid ** ****** *** many ********** *** ****; these ******* **** ** plan *** **********, *** the ********* *****, *** write ** *** ****** or ********* *********** *** results. ******** ***** *** those ********** ***** **** accrue **** **** *** additional ********* ** ****; they ******* *** ****** salary ** **** ********* as **** ** *** cost ** ********* *** evaluator's ****** *** *** cost ** *** ******** or ***** ********* **** during *** ********** *******. Based ** ********* ****** from ******* ******** *** fixed **** ** * heuristic ********** ** ********* to ** ******* $*,*** and $*,*** *** *** variable **** ** **** evaluator ** ********* ** be ******* $*** *** $900.

*** ****** ***** *** variable ***** **** ********* vary **** ******* ** project *** **** ****** on **** *******'* **** structure *** ** *** complexity ** *** ********* being *********. *** ************, consider * ****** ******* with ***** ***** *** heuristic ********** ** $*,*** and ******** ***** ** $600 *** *********. ** this *******, *** **** of ***** ********* ********** with*********** ** ****$(*,*** + ****).

*** ******** **** ********* evaluation *** ****** *** to *** ******* ** usability ********, ****** **** continuing ********* ******** *** be ******** ** *** extent **** *** ********** increase ***** ************* ** usability ** ********* ***** own ********** ******* **** those ** ***** **********. For **** ****** *******, assume **** ** ** worth $**,*** ** **** each ********* *******, ***** a ***** ******* ** Nielsen *** ******** (****) from ******* ********* *******. For **** ********, *** would ********* **** ** estimate *** ***** ** finding ********* ******** ***** on *** ******** **** population. *** ******** ** be **** **-*****, **** value *** ** ********* based ** *** ******** increase ** **** ************; for ******** ** ** sold ** *** **** market, ** *** ** estimated ***** ** *** expected ******** ** ***** due ** ****** **** satisfaction ** ****** ****** ratings. **** **** **** value **** ******* **** those ********* ******** **** are ** **** ***** before *** ******** *****. Since ** ** ********** to *** *** ********* problems, *** ***** ** each ******* ***** ** only **** ********** ** the ***** ** * fixed *******.

Figure * ***** ******* *** **** times *** ******** *** greater **** *** ***** for ********* ********** ** a ****** ******* ***** the *********** ********* ** the ****. *** ******* number ** ********** ** this ******* ** ****, with ******** **** *** 62 ***** ******* **** the *****.
IPVM Image

****** * ***** *** varying ***** ** *** benefits ** *** ***** for ******* ******* ** evaluators ** *** ****** project. *** ***** ***** that *** ******* ****** of ********** ** **** example ** ****, ********** the ******* *********** **** heuristic ********** ***** ** work **** **** ***** to **** **********. ** the *******, * ********* evaluation **** **** ********** would **** $*,*** *** would **** ********* ******** worth $***,***.

**********

  • *******, *. *. ****.* ********** ** ********* Evaluation *** ********* *******: The ******** ** * Domain-Specific ********* *********. **.*. ****., ********** of ********** ***********, ***** A&M **********, ******* *******, TX.
  • ********, *., ******, *. R., *******, *., *** Uyeda, *. *. ****. User ********* ********** ** the **** *****: * comparison ** **** **********.*********** *** ***'** **********(*** *******, **, ***** 28-May *), ***-***.
  • ******, *., *** *******, J. (****). ********* * human-computer ********,************** ** *** ***33 , 3 (March), 338-348.
  • *******, *. ****. ***** versus ******** *************** ** mockup ********* *** ********* evaluation.****. **** ********** ***** Intl. ****. *****-******** ***********(*********, *.*., ****** **-**), 315-320.
  • *******, *., *** ********, T. *. ****. * mathematical ***** ** *** finding ** ********* ********.*********** ***/**** ********'** **********(*********, *** ***********, ***** 24-29), ***-***.
  • *******, *., *** ******, R. (****). ********* ********** of **** **********,****. *** ***'** ****.(*******, **, *-* *****), 249-256.
  • *******, *. ****. ******* usability ******** ******* ********* evaluation.*********** *** ***'** **********(********, **, *** *-*), 373-380.
  • *******, *. (****). ********* evaluation. ** *******, *., and ****, *.*. (***.),********* ********** *******. **** ***** & Sons, *** ****, **.
Agree
Disagree
Informative
Unhelpful
Funny

**** ******* ** ********** the ********* ********* *** be **** ** **** to *** **** ** that '****** ********* **' systems **** **** *** some ** ******* ******* end **** ******* ** fact ** ***** ** in **** *** ******* neural ******* ** ** is **** ****** **********...

Agree: 1
Disagree
Informative
Unhelpful
Funny

* **** *********** ****. I ***** **** **** whether ** **** ****** the ****** **** ********** with *** ******* *****.

Agree
Disagree
Informative
Unhelpful
Funny

*****, ****** *** *** feedback. ***** ** ** not ******* ****** **** for *** ******** **** we ***** ** **** Fundamentals *****, ***** *** numerous ****-****** **** *** object ***********.

***** *** **** * few ******** ** ******** of ****-****** ********:

Agree
Disagree
Informative
Unhelpful
Funny

**** ************ ********! ** many ****** * ****'* know **** * ****'* know.

Agree: 1
Disagree
Informative: 1
Unhelpful
Funny

**** ** **** ******** and ******** ****** ***********.

Agree
Disagree
Informative: 1
Unhelpful
Funny
Read this IPVM report for free.

This article is part of IPVM's 7,250 reports and 966 tests and is only available to subscribers. To get a one-time preview of our work, enter your work email to access the full article.

Already a subscriber? Login here | Join now
Loading Related Reports