• Katie Read

Assessing algorithm and human facial composite identification decisions



School of Human Sciences

Institute of Lifecourse Development

University of Greenwich

London SE10 9LS

www.superrecognisers.com

super-recognisers@greenwich.ac.uk

Twitter:@GRecognisers

https://orcid.org/0000-0003-0017-7159

https://www.gre.ac.uk/pg/eduhea/psych

E2ID: Assessing algorithm and human facial composite identification decisions

Please note: This is a different E2ID study from the "Beat the Computer" research in which participants provided similarity ratings to 50 pairs of composite-photo images. You will find preliminary results from that study in a blog published on 6thAugust 2020.

https://www.superrecognisers.com/post/scores-of-first-10-000-to-complete-e2id-project

Research Team: 

Dr Josh P. Davis (Professor, University of Greenwich). Research Assistant: Nikolay Petrov (University of Greenwich)

Co-Investigator: Dr Elena Belanova (University of Greenwich)

Co-Investigators: VisionMetric team (Chistopher Solomon, Sivapriyaa Kannappan, Stuart Gibson, Yan Wang, Fangliang Bai)

Thanks to everyone for your help with this collaboration between the University of Greenwich and VisionMetric funded by Innovate UK (https://gtr.ukri.org/projects?ref=105042).Your participation is very valuable to us. Our research would not be possible without your help.

In three experiments, conducted from May to September 2020, we asked for help with a project attempting to improve face matching performance by combining algorithm and human decision making. If you contributed to this project you should find your scores below. Your results may differ substantially from your previous scores recorded on facial recognition research. Pease do not be disheartened. This test was designed to be very difficult (particularly Experiment 2), and participants were assigned to different conditions within experiments. Some conditions were harder than others. Your condition may have been particularly hard.

We aim to submit a paper for publication before the end of the year, so until then we cannot disclose full details of the project. We will upload a pre-print for view on the OSF system and will make an announcement. However, we know our participants like to see how they performed in projects, so we have published this blog listing the scores and overall performances.

Assessing face matching performance - Experiment 1 (28thof May 2020 – 18thof June 2020)

In the first experiment, participants completed 50 randomly allocated trials. In each trial, they viewed 16 images in a 4 x 4 array. Each array contained a facial composite of the target in the bottom-left corner, while the other 15 images were the randomly ordered photo of the target (ground truth image) and 14 foils selected by a computer algorithm as being of similar appearance to the facial composite. The system was designed so that before they could start, participants were required to adjust the size of the array on the screen (using zoom settings) so that all 16 images were visible simultaneously. In each trial, participants were required to select one of the 15 images that they believed matched the facial composite – see Figure 1.

Figure 1

Frequency of scores on the array face matching task for Experiment 1 (n= 140)



Assessing face matching performance - Experiment 2 (19thof June 2020 – 5thof August 2020)

The array matching task in Experiment 2 was virtually identical to Experiment 1 except only the hardest 25 trials from Experiment 1 were retained. These were presented in a random order. See Figure 2.

Figure 2

Frequency of scores on the array face matching task for Experiment 2 (n= 318)



Assessing face matching performance - Experiment 3 (26thof August 2020 – 23rdof September 2020)

The array face matching task mimicked those in Experiments 1 and 2 with a few changes. Each array contained the facial composite of the target in the bottom-left corner, the ‘best match’ image in the top-left corner (this image varied, see Design when the paper is published), while the other 14 images contained the randomly ordered 14 foils selected by a computer algorithm as being the best matches to the composite. Participants were required to select one of the 15 images that they believed matched the facial composite and provide a confidence rating on a scale of 1 (guessing) to 100 (absolutely certain). There was a total of 45 trials. Total scores are displayed in Figure 3.

Figure 3

Frequency of scores on the array face matching task for Experiment 3 (n= 264)



The scores - in participant code order can be found below. Note - these are the codes entered into the system. One or two did not match with our database.

Thanks

Josh Davis and Nikolay Petrov

Results for Experiment 1 (out of 50)

Code Score

A00010182-28

A00015933-18

A00027927-15

A00036226-16

A00041479-29

A00046191-32

A00046297-35

A00046358-31

A00046599-25

A00046985-36

A00046989-21

A00047107-43

A00047117-28

A00047545-11

A00047586-39

A00047814-31

A00047828-36

A00047915-35

A00047936-23

A00047953-40

A00047971-30

A00047974-30

A00048033-21

A00048116-42

A00048294-41

A00048331-24

A00048372-18

A00048379-24

A00048385-34

A00048439-30

A00048511-39

A00048765-23

A00048868-20

A00049100-14

A00049377-40

A00049381-34

A00049422-28

A00049442-26

A00049444-24

A00049635-30

A00050118-29

A00050268-24

A00050407-28

A00050407-32

A00050474-17

A00050754-42

A00050769-33

A00050866-29

A00050977-21

A00051186-39

A00051374-37

A00051444-28

A00051474-38

A00051476-34

A00051554-20

A00051590-17

A00051608-32

A00051662-32

A00051678-18

A00051699-36

A00051853-25

A00051896-31

A00051897-37

A00051980-19

A00052116-30

A00053156-21

A00053303-38

A00053344-39

A00053364-25

A00053375-40

A00053377-30

A00053451-31

A00053492-32

A00053825-23

A00053857-38

A00055341-28

A00055692-29

A00055754-39

A00057152-32

A00057581-35

A00058001-29

A00058024-29

A00058033-36

A00058936-29

A00058936-36

A00060390-33

A00060407-21

A00060441-34

A00060475-25

A00060479-35

A00060546-41

A00060589-36

A00060664-36

A00060676-36

A00060924-35

A00060948-29

A00061006-25

A00061314-32

A00061593-32

A00061610-34

A00061625-17

A00061706-37

A00061884-24

A00061939-40

A00061940-37

A00061995-20

A00062070-30

A00062118-30

A00062320-33

A00062532-32

A00062547-23

A00062831-32

A00062873-37

A00063107-37

A00063189-23

A00063242-31

A00068131-28

A00073926-31

A00113141-39

A00113770-28

A0011377438

A00113820-31

A00131337-17

A00207363-31

A00224963-37

A00261267-32

A00284021-33

A00285208-37

A00310681-35

A00313072-24

A00313563-21

A00313563-16

A00314241-25

A00315161-38

B00000120-35

B00003195-29

B00004237-21

B00006598-33

B00006818-35

T89239701-21

Results for Experiment 2 (out of 25):

Code Score

A00048493-12

A00048645-9

A00050286-4

A00051036-6

A00052288-5

A00052353-3

A00053221-9

A00057542-7

A00060771-10

A00060816-4

A00061684-7

A00062344-5

A00062375-8

A00062454-8

A00062941-5

A00064065-7

A00070346-6

A00084966-8

A00085040-5

A00089665-11

A00091996-9

A00097566-6

A00103834-9

A00113617-10

A00113844-10

A00115093-8

A00132721-8

A00133674-8

A00136930-14

A00138866-4

A00139054-7

A00142508-10

A00144879-4

A00146091-5

A00147072-6

A00147168-11

A00160070-6

A00221667-2

A00221899-8

A00222176-10

A00222342-9

A00225244-6

A00227057-9

A00228098-11

A00231112-11

A00236142-2

A00236373-5

A00241218-10

A00259925-6

A00259967-10

A00276310-3

A00277414-10

A00278110-4

A00280304-8

A00282804-4

A00283088-5

A00286169-2

A00287011-12

A00287452-14

A00287488-4

A00287557-4

A00287615-9

A00287618-7

A00287632-3

A00287972-5

A00288018-7

A00288040-8

A00290149-7

A00290598-5

A00290609-9

A00290953-6

A00290982-8

A00291104-6

A00292015-6

A00293955-9

A00294911-7

A00296241-2

A00297084-8

A00302952-5

A00303225-5

A00304246-8

A00304470-2

A00309336-6

A00309456-7

A00309542-6

A00309544-9

A00309562-14

A00309925-7

A00309926-6

A00309939-6

A00309944-7

A00309988-7

A00310006-9

A00310080-8

A00310105-6

A00310146-3

A00310170-3

A00310177-3

A00310186-10

A00310198-8

A00310210-9

A00310215-6

A00310272-5

A00310286-2

A00310310-11

A00310340-8

A00310348-6

A00310478-9

A00310480-11

A00310583-4

A00310669-7

A00310673-4

A00310692-6

A00310776-8

A00310977-1

A00311053-8

A00311101-9

A00311254-10

A00311485-11

A00311841-5

A00312870-8

A00313238-3

A00313243-5

A00313365-9

A00313520-10

A00313674-8

A00313863-10

A00313871-2

A00313880-7

A00313881-6

A00313925-6

A00314026-6

A00314045-10

A00314072-5

A00314155-12

A00314190-6

A00314224-5

A00314402-2

A00314462-8

A00314471-5

A00314593-5

A00314736-9

A00314756-8

A00314770-0

A00314844-6

A00314987-6

A00314994-7

A00314999-6

A00315063-4

A00315109-9

A00315110-6

A00315117-12

A00315141-9

A00315215-9

A00315299-6

A00315344-12

A00315374-5

A00315578-5

A00315651-8

A00315660-2

A00315705-6

A00315787-1

A00315945-5

A00316078-8

A00316261-12

A00316300-5

A00316317-8

A00316378-6

A00316394-6

A00316422-5

A00316426-3

A00316468-5

A00316526-8

A00316537-10

A00316789-6

A00316807-4

A00316853-11

A00316919-6

A00317017-2

A00317144-5

A00317318-10

A00317328-11

A00317330-7

A00317335-3

A00317370-4

A11111111-9

B00002575-7

B00002615-9

B00002750-5

B00002866-5

B00002891-10

B00003000-14

B00003001-8

B00003078-11

B00003169-13

B00003171-13

B00003177-7

B00003307-8

B00003368-7

B00003378-6

B00003423-5

B00003638-4

B00003674-5

B00003726-8

B00003730-7

B00003765-3

B00003788-9

B00003789-6

B00004036-7

B00004193-5

B00004195-6

B00004281-10

B00004282-1

B00004340-2

B00004765-4

B00004823-6

B00004895-7

B00004907-5

B00005117-6

B00005134-4

B00005185-4

B00005256-11

B00005258-4

B00005292-9

B00005462-3

B00005509-8

B00005566-9

B00005867-7

B00005887-6

B00006045-8

B00006061-10

B00006085-11

B00006188-5

B00006320-7

B00006367-9

B00006736-5

B00006781-7

B00006782-7

B00006862-6

B00007012-4

B00007081-4

B00007083-14

B00007130-8

B00007142-2

B00007302-5

B00007333-5

B00007367-3

B00007400-10

B00007483-6

B00007760-4

B00007846-3

B00007986-11

B00008024-4

B00008068-7

B00008088-5

B00008094-7

B00008106-7

B00008122-8

B00008197-3

B00008228-4

B00008267-8

B00008286-13

B00008470-5

B00010510-5

B00010526-7

B00010561-8

B00010562-9

B00010597-8

B88888888-10

M98798798-6

N00000000-3

T00531405-6

T00992563-1

T07227843-13

T08348918-5

T09779640-10

T11055834-3

T13206264-10

T13223566-4

T13317154-10

T13631697-2

T17018898-3

T18855185-12

T22928008-5

T26197609-8

T26408039-6

T29212079-12

T30127199-9

T41155523-9

T46141083-6

T46669602-4

T47261427-5

T50263705-4

T51083841-7

T55247412-6

T56272142-3

T59644452-10

T68503103-9

T75737600-6

T76746853-3

T76810589-18

T79381895-1

T81019073-8

T82639714-10

T84867634-10

T84900167-8

T86434926-12

T86974216-5

T87023835-9

T87023835-10

T92073820-4

T92508266-5

T92745577-3

T93730730-13

T94238324-12

T99849024-11

T99867585-11

V60000009-13


Results for Experiment 3 (out of 45):

Code Score

A00001024-21

A00001640-37

A00001723-26

A00002162-33

A00003333-30

A00003392-25

A00003737-19

A00003760-27

A00003767-24

A00004005-13

A00004856-12

A00006354-28

A00007056-25

A00011351-20

A00011370-25

A00011787-10

A00015373-27

A00015551-23

A00017624-30

A00020957-0

A00031520-16

A00035414-35

A00035695-26

A00035990-27

A00036217-22

A00038496-28

A00038500-36

A00038964-25

A00040286-27

A00041240-22

A00041302-30

A00041409-26

A00041427-33

A00041449-21

A00041612-29

A00041790-24

A00042778-22

A00044690-36

A00045909-29

A00048768-21

A00051474-24

A00051474-29

A00053221-31

A00058853-24

A00060634-25

A00060816-36

A00060869-27

A00060942-32

A00060968-39

A00061675-29

A00062375-28

A00064551-32

A00065020-30

A00067205-29

A00067792-25

A00069874-28

A00070149-34

A00070346-27

A00071446-26

A00071751-24

A00086145-25

A00087407-37

A00096952-31

A00099449-0

A00105332-37

A00105441-27

A00113310-14

A00113430-20

A00113653-29

A00113743-30

A00114237-42

A00114287-30

A00114580-29

A00114890-16

A00115050-33

A00115197-20

A00115200-35

A00115286-29

A00115477-34

A00115579-27

A00121853-30

A00132051-21

A00132092-25

A00132768-28

A00132887-33

A00133650-38

A00133674-26

A00134098-34

A00134128-10

A00137739-31

A00138436-23

A00138533-28

A00138866-23

A00141974-36

A00142653-31

A00142863-21

A00143460-34

A00143646-35

A00145530-28

A00145595-20

A00145644-33

A00146091-22

A00146128-25

A00146969-35

A00148157-29

A00159540-35

A00166325-22

A00167470-29

A00167994-26

A00170235-23

A00172595-26

A00172650-18

A00180508-26

A00213662-23

A00218077-16

A00221230-16

A00221709-26

A00222326-23

A00222752-21

A00223082-33

A00223258-34

A00223723-35

A00223830-24

A00223973-24

A00224178-0

A00224307-17

A00224409-0

A00224552-17

A00224773-35

A00224886-30

A00226130-25

A00227051-22

A00228668-15

A00228803-35

A00229045-14

A00230064-32

A00230435-25

A00231722-18

A00231729-18

A00240677-14

A00240677-29

A00241012-25

A00259172-35

A00260117-21

A00260967-25

A00275857-27

A00275954-36

A00276094-27

A00276122-20

A00276171-28

A00277682-29

A00277793-34

A00279224-28

A00281805-28

A00284648-19

A00285749-0

A00286964-18

A00292010-18

A00292131-23

A00292570-21

A00295718-25

A00297181-27

A00297332-33

A00302859-17

A00303644-33

A00303816-22

A00309434-33

A00309542-33

A00309573-13

A00309599-22

A00309826-24

A00310066-21

A00310086-42

A00310238-26

A00310241-31

A00310273-27

A00310406-22

A00310586-29

A00310674-38

A00310760-32

A00310789-37

A00310855-37

A00311005-23

A00311141-33

A00311328-23

A00311390-34

A00311437-35

A00311485-29

A00312618-28

A00313098-25

A00313394-8

A00313494-23

A00313534-33

A00313757-23

A00313809-34

A00313820-30

A00314837-24

A00314995-26

A00315177-21

A00315297-24

A00315369-29

A00315466-21

A00315715-13

A00315777-11

A00315894-32

A00315919-31

A00315934-34

A00316388-24

A00316627-25

A00316864-31

A00316919-32

A00316970-19

A00316990-23

A00317030-26

A00317436-10

A00317484-19

B00000030-23

B00000080-31

B00000197-30

B00000279-28

B00000409-22

B00002336-34

B00002817-28

B00002859-27

B00003075-17

B00003140-30

B00003149-25

B00003166-31

B00003168-35

B00003179-24

B00003212-32

B00003339-24

B00006473-12

B00006532-18

B00006770-0

B00007215-29

B00007709-25

B00007800-26

B00008189-19

B00008210-13

B00008305-19

B00008306-22

B00008342-17

T10518348-1

T11263536-19

T22783766-6

T25015964-15

T27305970-24

T28120793-15

T31206026-27

T33630779-34

T34035959-40

T47456966-6

T57315921-28

T65096833-28

T75625785-26

T78029271-17

T81466364-18

T81802522-19

T88914748-29

T89433824-31

T92718638-31

T92772632-26

T93625585-31



172 views

Recent Posts

See All
  • Email icon

Where to find us?

Greenwich Super-Recognisers

Supervisor: Dr Josh P Davis

Applied Psychology Research Group

University of Greenwich

Old Royal Naval College,

Park Row

London

SE10 9LS

©2019 by Super-Recognisers.