Research Article

Do Deepfakes Adequately Display Emotions? A Study on Deepfake Facial Emotion Expression

Table 4

Sample data and detected overall emotions.

IDORPFFRPDynamic modelStatic model
COORFA%CO%OR%FACOORFA%CO%OR%FA

id0101311426082406302652.0147.9960.3550298427701237.3762.6352.11
id1101715389438529640551.1848.8236.6612674111171246953.2746.7352.41
id2101715816521871386646.8953.11109.74551510780825333.8466.1650.65
id3101715738686988269135.6364.3724.79440942801658550.7449.26190.87
id4101514039394129300148.8251.1837.21166579091167759.5940.4159.66
id51031850845245352.9247.0847.192323156895959.7040.3024.65
id6101713437183055338654.8945.1149.99154878694776064.0535.9532.09
id7104291167779117059.9740.0360.1234191811247965.3734.6347.40
id8109332990298066150.0849.9211.0734864096262645.9854.0234.63
id9101712455824460265255.5944.4126.4188509186389249.0750.9321.58
id101043331551101108974.1325.8725.593376242878558.1741.8313.53
id111111187735623871.1328.8719.3168692671964.5535.4527.53
id12721194122954380.4319.5746.41155957692873.0226.9843.47
id13162232169126346563.236.813.553422290497454.0945.9115.40
id16141620358538438379240.9659.0426.5323137197101457454.0046.0034.01
id17101614384484272177366.4233.5813.94104276117531163.0336.9732.10
id19109722342951166771.1228.8850.6232343449285848.3951.6142.77
id20102721271946925344350.9549.0524.393372430693839852.3547.6513.04
id21102720514863504250829.7870.2250.2612585104291084054.6845.3247.10
id22109529613961540.8559.15261.743713393141556.3043.7018.23
id23102721853535838424647.8352.1737.9424505161981387860.2039.8034.10
id241097944022536228063.4536.5532.8665263265501366.6533.3551.20
id251195915841279106755.3344.6737.2735965512235239.4860.5225.82
id26102717050755278250149.0250.9824.16698849531156158.5241.4896.82
id27109772147251787846.0353.9718.8329511430324767.3632.6474.12
id28102724932363125543850.8749.1385.4915921101561314961.0538.9550.42
id29101310329013679172244.0955.9126.171021110678629948.8851.1230.15
id30102120127313096777646.8753.13133.454063839511942750.7049.3011.76
id31102119713724774356622.3277.6858.0239716300331602356.9443.0622.97
id3210138814351081216357.0342.9785.97160326920559369.8530.1524.37
id33101310942173536485254.3945.6162.5818643157521170254.2045.8034.02
id3410138746542113316268.7731.2346.7372655690522256.0843.9240.31
id35102118246535571536045.5154.4952.4315156148801232350.4649.5441.03
id37102117231973374526448.6551.3580.1132742214841662960.3839.6230.67
id3810131043032295779050.6349.3713.1980025657729558.5841.4253.41
id391087236941242162074.8425.1632.82107105169500567.4532.5531.52
id401098375272201178377.3722.6318.33123274566418972.9727.0324.80
id411087134972381161659.4940.5127.4993805408400263.4336.5727.06
id4259363505968118478.3621.6426.4739941109109578.2721.7321.46
id431097954431088376183.3416.6657.5997518985437152.0447.9623.33
id44694618611435101356.4643.5430.7328202540113952.6147.3921.25
id451098734113814240847.2152.7933.3342964011364051.7248.2843.82
id461097619451702157353.3346.6743.1332333416299348.6251.3845.01
id471098938841660242670.0629.9443.76144754996622174.3425.6631.95
id481098450352619232265.7834.2230.34104345901467863.8836.1228.64
id491098424061865389456.3343.6791.1790718312501152.1847.8228.83
id501087133023216447850.6649.3468.748212292826567.7832.22116.20
id5110867112244040185073.5326.4712.12171907082443170.8229.1818.26
id521087989316069281259.5440.4618.752133214171628560.0939.9117.70
id531098698106115255761.638.416.061729510690370761.8038.2013.25
id5410980117345870310666.6633.3417.64137036736669467.0432.9632.75
id551096475762783439573.1326.8742.43108374537574370.4929.5137.36
id561097991313168336674.2425.7627.37149687330580467.1332.8726.03
id571096348742652336664.7635.2444.7268965005373557.9442.0631.38
id581085544362726320861.9438.0644.7961224539417957.4242.5839.20
id5910323103684095555.2244.7850.9130381975273560.6039.4054.56
id601033018622647168241.358.737.338534919341543.9256.0838.93
id611031536048390742.757.3107.5917961459128555.1844.8239.48

Columns display the number of original recordings by the performer (OR), the number of performers of deepfakes on the original recordings of the current performer (PF), and number of fake recordings made on recordings by the performer (FRP). Then, the number of photograms with emotions in common with the original recording and the corresponding photogram of a fake recording (CO), the number of photograms that showed emotion in the original recording but not in the corresponding fakes (OR), and the number of emotional photograms that appear in fakes but not in corresponding photograms on original recordings alone (FA) are displayed followed by their corresponding percentages, for dynamic and static models.