Pick female hockey players

**~~Komintasavalta~~** · 11-27-2020, 07:09 PM

Originally Posted by hecate

1,5,9,11,15,20,24,38,40,52,59,63,67,68,69,74,80,95 ,97,104,110,115,124,125,135,139,146,151,161,169

Your morph ended up looking pretty boring and average too, even though it has darker pigmentation than average:

You're another solid member of the wog cluster. You have very low relatedness with most members of the Uralic cluster. You have the highest relatedness with users in the "Intermediate types" and "Darkest pigmentation" groups:

.59;IceQueen;Intermediate types
.57;stellan;Intermediate types
.55;Parça do Neymar;Darkest pigmentation
.53;Celestia;Darkest pigmentation
.52;akva;Intermediate types
.50;itilvolga;Small and short face
.50;Tacitus;Intermediate types
.46;Methuselah;Uralic
.45;Nelju;Intermediate types
.44;Helen;Small and short face
.44;Insuperable;Nordid
.43;RN97;Intermediate types
.43;hmaohma78;Intermediate types
.42;Kyp;Small and short face
.41;IrisSelene;Uralic
.40;placebo;Intermediate types
.40;pajkosbalna;Intermediate types
.40;Not a Cop;Intermediate types
.40;KirillMazur;Intermediate types
.40;Defcon2;Intermediate types
.39;CommonSense;Intermediate types
.37;Peterski;Intermediate types
.37;Ylla;Medium-sized round face
.37;calxpal;Darkest pigmentation
.36;Creoda;Nordid
.36;Sandis;Robust and high face
.35;Mixedguy17;Darkest pigmentation
.34;Lozano;Darkest pigmentation
.34;Dan Coren;Nordid
.34;Bmack;Nordid
.33;Spaniard 2.0;Small and short face
.30;PaleoEuropean;Medium-sized round face
.30;Yesno;Darkest pigmentation
.28;Immanenz;Robust and high face
.27;XenophobicPrussian;Nordid
.27;RatCat;Small and short face
.27;Lemminkäinen;Darkest pigmentation
.27;Alenka;Large round face
.26;TheMaestro;Darkest pigmentation
.26;Stearsolina;Uralic
.26;mergen3;Small and short face
.24;Thracian;Nordid
.24;Benyzero;Small and short face
.24;2garin;Nordid
.23;Aeduard;Intermediate types
.22;WeirdLookingFellow;Large round face
.21;Óttar;Darkest pigmentation
.21;Vid Flumina;Darkest pigmentation
.17;Aodhan;Large round face
.16;Hektor12;Uralic
.16;War Chef;Intermediate types
.13;ChineseKurd;Uralic
.12;Yamyam;Uralic
.12;Teutone;Darkest pigmentation
.12;Hamilcar;Darkest pigmentation
.10;Edgü;Uralic
.10;travv;Uralic
.10;Komintasavalta;Uralic
.09;ZzeroKrieg;Nordid
.08;Malagueña;Uralic
.06;cass;Large round face

**IrisSelene** · 11-27-2020, 10:45 PM

I think most peoples morph tend to look like themselves a bit lol

Also, its a bit unfair bc of course the morphs of those who choose less than 20 are gonna look prettier and unique, cuz less faces are mixed in, when you pass that number of people they tend to blend in together too much to rlly tell apart features easily, imo.

Enviado desde mi CLT-L09 mediante Tapatalk

**~~Komintasavalta~~** · 11-29-2020, 02:14 PM

Originally Posted by IrisSelene

I think most peoples morph tend to look like themselves a bit lol

Yeah, a good example is Nurzat's morph in the other thread:

Originally Posted by IrisSelene

I think most peoples morph tend to look like themselves a bit lolAlso, its a bit unfair bc of course the morphs of those who choose less than 20 are gonna look prettier and unique, cuz less faces are mixed in, when you pass that number of people they tend to blend in together too much to rlly tell apart features easily, imo.

Morphs with more photos are supposed to be more attractive though (https://en.wikipedia.org/wiki/Averageness):

In 1990, one of the first computer-based photographic attractiveness rating studies was conducted. During this year psychologists Langlois and Roggman wanted to systematically examine whether mathematical averageness is linked with facial attractiveness.[1][2][3][11][12][13][14] To test this, they selected photographs of 192 young male and female Caucasian faces; each of which was computer scanned and digitized. They then made computer-processed composites of each image, as 2-, 4-, 8-, 16-, and 32-face composites, averaged by pixel. These faces, as well as the component faces, were rated for attractiveness by 300 judges on a 5-point Likert scale (1 = very unattractive, 5 = very attractive). The 32-composite face was the most visually attractive of all the faces.[1]

Here's morphs of randomly selected players. I think maybe the morphs with 64 or 32 players look the most attractive on average.

Actually maybe next I'll make a thread where I'll make users rate morphs of randomly selected people, and then I derive the attractiveness of the people from the rating of the morphs they are part of.

**~~Komintasavalta~~** · 11-29-2020, 08:59 PM

In case someone still wants to vote, I created this HTML interface you can use to select the images more easily: https://output.jsbin.com/wurijap.

Try to select about 30 best picks. You can optionally include about 30 worst picks. Exactly 30 is even better.

**~~barnumandbailey~~** · 11-29-2020, 11:18 PM

I managed to pick twenty two girls out, and have listed them in numerical order below:

1 5 6 8 11 13 15 26 40 44 47 59 63 65 97 98 115 125 139 161 192 200

Out of those, there were only a few who I'd actually consider "attractive", but I went ahead and picked the women I found to be "prettiest". Hope this helps!

**~~Komintasavalta~~** · 11-29-2020, 11:26 PM

Originally Posted by barnumandbailey

I managed to pick twenty two girls out, and have listed them in numerical order below:

1 5 6 8 11 13 15 26 40 44 47 59 63 65 97 98 115 125 139 161 192 200

Out of those, there were only a few who I'd actually consider "attractive", but I went ahead and picked the women I found to be "prettiest". Hope this helps!

I thought at first that your morph looked like the morphs of Sandis and Immanenz, who are in the "high and robust face" category. But actually your face is not as high though:

Also you have highest relatedness with users in the "intermediate types" and "darkest pigmentation" groups:

Spoiler!

**savvas** · 12-04-2020, 05:21 PM

14, 15, 20, 29, 42, 44, 52, 67, 68, 78, 82, 95, 97, 99, 100, 104, 110, 111, 114, 115, 167, 168, 169, 186, 190, 192, 196, 201, 204, 205

Spoiler!

**~~Komintasavalta~~** · 12-04-2020, 06:03 PM

Originally Posted by savvas

14, 15, 20, 29, 42, 44, 52, 67, 68, 78, 82, 95, 97, 99, 100, 104, 110, 111, 114, 115, 167, 168, 169, 186, 190, 192, 196, 201, 204, 205

Spoiler!

I didn't figure out what cluster I should assign your morph in. The users who had the highest relatedness with you were from multiple different clusters:

Spoiler!

I had earlier rated all players in this thread on a six-point scale, with this distribution: Very best (10), Best (20), Better (34), Average (69), Worse (42), Worst (30).

Here's your top 30 picks grouped by my rating for them:

Compared to the last morph, the first morph has lower upper eyelid exposure, higher canthal tilt, a shorter distance between the eyebrows and the eyes, bigger cheekbones, bigger teeth, and a wider nose. It also has a lower y-axis position of the glabella, and its eyebrows are less slanted outwards.

**~~Komintasavalta~~** · 12-04-2020, 06:04 PM

double

**~~Komintasavalta~~** · 12-05-2020, 03:28 AM

I used Google's Cloud Vision API to get the coordinates of facial landmarks in the photos in this thread. You can try a demo here: https://cloud.google.com/vision/docs/drag-and-drop.

I ran a shell command like this for each headshot photo in this thread:

curl -HAuthorization:Bearer\ `gcloud auth application-default print-access-token` -HContent-Type:application/json\;charset=utf-8 https:/\/vision.googleapis.com/v1/images:annotate -d@<(printf %s '{"requests":[{"image":{"content":"'"`base64 input.jpg`"'"},"features":[{"maxResults":1,"type":"FACE_DETECTION"}]}]}')

Then for example I ran a command like this to calculate the distance between the left and right eyes:

jq -r '[.responses[0].faceAnnotations[].landmarks[]|select(.type=="LEFT_EYE"or.type=="RIGHT_EYE")|.po sition|[.x,.y,.z]]|[transpose[]]|map(.[1]-.[0]|.*.)|add|sqrt'

Based on the landmarks returned by the Cloud Vision API, I generated various metrics. I converted the values of each metric into z-scores by subtracting the average value and dividing by the standard deviation.

In each pair of morphs below, the morph on the left side consists of the 16 players with the highest value for the metric, and the morph on the right side consists of the 16 players with the lowest value for the metric. I didn't exclude players who were smiling with their mouth open, which caused many metrics to be calculated incorrectly, including the thickness of lips, the width of the mouth, and the distance between nose and mouth. All of the metrics are to some extent also affected by the angle, distance, or focal length of the camera. The angle of the photo has a particularly large effect on canthal tilt, as you can see from the morphs below.

Here's a table of the z-scores for each player:

Spoiler!

I also used this tool to create a PCA chart based on everyone's z-scores: https://biit.cs.ut.ee/clustvis/. The chart is based only on the picks for best players. I ignored users with less than 10 total picks.

Malagueña had a low number of shared picks with almost everyone, so she was expectedly an outlier. However she plots the closest to travv and Edgü, whose morphs also looked similar to her morph.

The more extreme members of the Uralic cluster like travv, me, and Edgü/Yamyam/Chinese Kurd plot further away from the main cluster. The more moderate members of the Uralic cluster like Stearsolina, IrisSelene, and Hektor12 plot closer to the main cluster.

Lemminkäinen is another outlier. He was the only user who picked the most players out of the French team. His picks have low average z-scores for nasal index and jaw width, and they have high z-scores for facial index, brow height, and palpebral fissure height.

The PCA chart above is based only on morphology and not on pigmentation, because it's based on the metrics that I generated from the coordinates of facial landmarks that were returned by Google's Cloud Vision API.