Validation results for the models inferring using Intel® Distribution of OpenVINO™ Toolkit

Object detection

Test image #1

Image resolution: 2048 x 1024

Bounding box (upper left and bottom right corners):
CAR (0, 431), (231, 914)
CAR (279, 430), (442, 519)
CAR (428, 444), (494, 498)
CAR (719, 431), (827, 519)
CAR (789, 405), (874, 493)
CAR (828, 413), (970, 536)
CAR (938, 417), (1021, 497)
CAR (1037, 428), (1069, 457)
CAR (1092, 413), (1196, 509)
PERSON (1455, 419), (1482, 491)
PERSON (1476, 416), (1503, 481)

Model	Python (latency mode, implementation)	Python (throughput mode, implementation)
pedestrian-and-vehicle-detector-adas-0001	Bounding box: CAR (720, 439), (821, 505), CAR (824, 424), (967, 525), CAR (945, 420), (1023, 486), CAR (1092, 422), (1188, 501), PERSON (1474, 416), (1499, 481)	Bounding box: CAR (720, 439), (821, 505), CAR (824, 424), (967, 525), CAR (945, 420), (1023, 486), CAR (1092, 422), (1188, 501), PERSON (1474, 416), (1499, 481)

Test image #2

Data source: Cityscapes

Image resolution: 2048 x 1024

Bounding boxes (upper left and bottom right corners):
CAR (360, 354), (917, 781)
CAR (906, 402), (1059, 522)
CAR (1175, 366), (1745, 497)
CAR (1245, 372), (1449, 504)
CAR (1300, 311), (1825, 605)
CAR (1599, 314), (2048, 625)
CAR (1697, 315), (2048, 681)

Model	Python (latency mode, implementation)	Python (throughput mode, implementation)
vehicle-detection-adas-0002	Bounding box: CAR (384, 363), (921, 754), CAR (909, 407), (1056, 509), CAR (1272, 348), (1742, 592), CAR (1618, 305), (2036, 669)	Bounding box: CAR (384, 363), (921, 754), CAR (909, 407), (1056, 509), CAR (1272, 348), (1742, 592), CAR (1618, 305), (2036, 669)
vehicle-detection-adas-binary-0001	Bounding box: CAR (370, 353), (905, 756), CAR (902, 406), (1048, 509), CAR (1246, 320), (2022, 650)	Bounding box: CAR (370, 353), (905, 756), CAR (902, 406), (1048, 509), CAR (1246, 320), (2022, 650)

Test image #3

Data source: Cityscapes

Image resolution: 2048 x 1024

Bounding boxes (upper left and bottom right corners):
CAR (0, 380), (88, 524)
CAR (107, 384), (327, 480)
CAR (506, 375), (623, 458)
CAR (626, 367), (734, 452)
CAR (919, 362), (968, 401)
CAR (1053, 360), (1091, 388)
BIKE (300, 402), (558, 778)
PERSON (310, 171), (536, 749)
PERSON (1779, 268), (1882, 539)
PERSON (1874, 288), (1976, 545)

Model	Python (latency mode, implementation)	Python (throughput mode, implementation)
person-vehicle-bike-detection-2000	Bounding box: CAR (121, 391) (318, 477) CAR (551, 374) (647, 456) CAR (634, 385) (722, 437) BIKE (310, 210) (551, 774) PERSON (1790, 284) (1885, 531) PERSON (1892, 301) (1977, 515)	Bounding box: CAR (121, 391) (318, 477) CAR (551, 374) (647, 456) CAR (634, 385) (722, 437) BIKE (310, 210) (551, 774) PERSON (1790, 284) (1885, 531) PERSON (1892, 301) (1977, 515)
person-vehicle-bike-detection-2001	Bounding box: CAR (108, 394) (310, 477) BIKE (321, 185) (545, 742) PERSON (1782, 278) (1889, 538) PERSON (1877, 295) (1971, 530)	Bounding box: CAR (108, 394) (310, 477) BIKE (321, 185) (545, 742) PERSON (1782, 278) (1889, 538) PERSON (1877, 295) (1971, 530)
person-vehicle-bike-detection-2002	Boounding box: CAR (554, 376) (666,455) CAR (0, 401) (86, 517) BIKE (334, 200) (540, 635) PERSON (1783, 269) (1881, 540) PERSON (1878, 298) (1973, 527) PERSON (329, 193) (545, 627)	Boounding box: CAR (554, 376) (666,455) CAR (0, 401) (86, 517) BIKE (334, 200) (540, 635) PERSON (1783, 269) (1881, 540) PERSON (1878, 298) (1973, 527) PERSON (329, 193) (545, 627)
person-vehicle-bike-detection-crossroad-0078	Bounding box: CAR (-4, 400), (80, 515), CAR (114, 392), (326, 480), CAR (547, 382), (645, 457), CAR (627, 379), (724, 444), BIKE (319, 232), (546, 717), PERSON (329, 228), (541, 697), PERSON (1783, 278), (1887, 530), PERSON (1882, 294), (1974, 524)	Bounding box: CAR (-4, 400), (80, 515), CAR (114, 392), (326, 480), CAR (547, 382), (645, 457), CAR(627, 379), (724, 444), BIKE(319, 232), (546, 717), PERSON (329, 228), (541, 697), PERSON (1783, 278), (1887, 530), PERSON (1882, 294), (1974, 524)
person-vehicle-bike-detection-crossroad-1016	Bounding box: CAR (-1, 405), (85, 518), CAR (533, 370), (637, 455), PERSON (319, 213), (554, 722), PERSON (1783, 270), (1884, 536), PERSON (1883, 299), (1975, 513)	Bounding box: CAR (-1, 405), (85, 518), CAR (533, 370), (637, 455), PERSON (319, 213), (554, 722), PERSON (1783, 270), (1884, 536), PERSON (1883, 299), (1975, 513)

Test image #4

Data source: GitHub

Image resolution: 799 x 637

Bounding boxes (upper left and bottom right corners):
CAR (232, 119), (509, 466)
PLATE (330, 410), (393, 436)

Model	Python (latency mode, implementation)	Python (throughput mode, implementation)
vehicle-license-plate-detection-barrier-0106	Bounding box: CAR (232, 119), (509, 466), PLATE (330, 410), (393, 436)	Bounding box: CAR (232, 119), (509, 466), PLATE (330, 410), (393, 436)

Test image #5

Data source: Internet

Image resolution: 320 x 320

Bounding boxes (upper left and bottom right corners):
PERSON (35, 17), (84, 192)
PERSON (79, 13), (122, 194)
PERSON (211, 78), (273, 279)

Model	Python (latency mode, implementation)	Python (throughput mode, implementation)
person-detection-asl-0001	Bounding box: PERSON (35, 17), (84, 192), PERSON (79, 13), (122, 194), PERSON (211, 78), (273, 279)	Bounding box: PERSON (35, 17), (84, 192), PERSON (79, 13), (122, 194), PERSON (211, 78), (273, 279)
person-detection-asl-0200	Bounding box: PERSON (28, 12), (66, 155), PERSON (63, 8), (97, 158), PERSON (169, 65), (215, 223)	Bounding box: PERSON (28, 12), (66, 155), PERSON (63, 8), (97, 158), PERSON (169, 65), (215, 223)
person-detection-asl-0201	Bounding box: PERSON (41, 19), (97, 235), PERSON (254, 96), (320, 335), PERSON (93, 15), (147, 236)	Bounding box: PERSON (41, 19), (97, 235), PERSON (254, 96), (320, 335), PERSON (93, 15), (147, 236)
person-detection-asl-0202	Bounding box: PERSON (338, 127), (430, 446), PERSON (127, 16), (195, 315), PERSON (56, 27), (129, 314)	Bounding box: PERSON (338, 127), (430, 446), PERSON (127, 16), (195, 315), PERSON (56, 27), (129, 314)

Test image #6

Data source: Internet

Image resolution: 512 x 512

Bounding boxes (upper left and bottom right corners):
PRINGLES (133, 195), (257, 195)
SPRITE (240, 487), (380, 10)

Model	Python (latency mode, implementation)	Python (throughput mode, implementation)
product-detection-0001	Bounding box: PRINGLES (130, 178), (275, 493)	Bounding box: PRINGLES (130, 178), (275, 493)

Test image #7

Data source: Wider Face

Image resolution: 1024 x 678

Bounding boxes (upper left and bottom right corners):
(189, 140) (288, 284)
(616, 45) (704, 213)

Model	Python (latency mode, implementation)	Python (throughput mode, implementation)
face-detection-0200	Bounding box: (188, 143) (284, 275), (616, 47) (700, 204)	Bounding box: (188, 143) (284, 275), (616, 47) (700, 204)
face-detection-0202	Bounding box: (189, 139) (285, 277), (613, 38) (701, 204)	Bounding box: (189, 139) (285, 277), (613, 38) (701, 204)
face-detection-0204	Bounding box: (189, 142) (288, 275), (614, 43) (704, 204)	Bounding box: (189, 142) (288, 275), (614, 43) (704, 204)
face-detection-adas-0001	Bounding box: (189, 140) (288, 284), (616, 45) (704, 213)	Bounding box: (189, 140) (288, 284), (616, 45) (704, 213)
face-detection-adas-binary-0001	Bounding box: (186, 137) (289, 277), (616, 53) (706, 211)	Bounding box: (186, 137) (289, 277), (616, 53) (706, 211)
face-detection-retail-0004	Bounding box: (189, 143) (286, 275), (613, 57) (694, 201)	Bounding box: (189, 143) (286, 275), (613, 57) (694, 201)
face-detection-retail-0005	Bounding box: (189, 140) (296, 277), (609, 44) (714, 206)	Bounding box: (189, 140) (296, 277), (609, 44) (714, 206)
face-detection-0100	Bounding box: (190, 142) (290, 282), (615, 46) (703, 210)	Bounding box: (190, 142) (290, 282), (615, 46) (703, 210)
face-detection-0102	Bounding box: (187, 141) (292, 280), (617, 50) (712, 210)	Bounding box: (187, 141) (292, 280), (617, 50) (712, 210)
face-detection-0104	Bounding box: (190, 142) (290, 280), (613, 43) (709, 211)	Bounding box: (190, 142) (290, 280), (613, 43) (709, 211)
face-detection-0105	Bounding box: (188, 141) (286, 279), (612, 45) (704, 204)	Bounding box: (188, 141) (286, 279), (612, 45) (704, 204)

Test image #8

Data source: Internet

Image resolution: 1999 x 1333

Bounding boxes (upper left and bottom right corners):
(1537, 385) (1792, 1184)
(541, 299) (845, 1161)
(229, 337) (453, 1048)
(0, 293) (193, 1129)
(955, 387) (1169, 1009)
(435, 370) (599, 1019)
(887, 292) (951, 479)
(749, 252) (866, 657)
(515, 317) (599, 580)
(833, 264) (894, 464)
(954, 283) (1020, 476)

Model	Python (latency mode, implementation)	Python (throughput mode, implementation)
person-detection-retail-0002	Bounding box: (252, 294) (465, 1048), (966, 361) (1183, 1028), (429, 262) (849, 1048), (695, 283) (872, 839), (421, 315) (612, 986), (1560, 360) (1766, 1204), (885, 283) (944, 503), (771, 276) (868, 574), (0, 314) (180, 941), (1879, 459) (1936, 694), (962, 279) (1023, 499), (1890, 302) (1992, 638)	Bounding box: (252, 294) (465, 1048), (966, 361) (1183, 1028), (429, 262) (849, 1048), (695, 283) (872, 839), (421, 315) (612, 986), (1560, 360) (1766, 1204), (885, 283) (944, 503), (771, 276) (868, 574), (0, 314) (180, 941), (1879, 459) (1936, 694), (962, 279) (1023, 499), (1890, 302) (1992, 638)
person-detection-retail-0013	Bounding box: (1537, 385) (1792, 1184), (541, 299) (845, 1161), (229, 337) (453, 1048), (0, 293) (193, 1129), (956, 387) (1169, 1009), (435, 370) (599, 1019), (887, 292) (951, 479), (749, 252) (866, 657), (515, 317) (599, 580), (833, 264) (894, 464), (954, 283) (1020, 476)	Bounding box: (1537, 385) (1792, 1184), (541, 299) (845, 1161), (229, 337) (453, 1048), (0, 293) (193, 1129), (956, 387) (1169, 1009), (435, 370) (599, 1019), (887, 292) (951, 479), (749, 252) (866, 657), (515, 317) (599, 580), (833, 264) (894, 464), (954, 283) (1020, 476)

Test image #9

Data source: Cityscapes

Image resolution: 1999 x 1333

Bounding boxes (upper left and bottom right corners):
(629, 310) (934, 811)
(392, 435) (440, 525)

Model	Python (latency mode, implementation)	Python (throughput mode, implementation)
pedestrian-detection-adas-0002	Bounding box: (614, 307) (945, 803)	Bounding box: (614, 307) (945, 803)
pedestrian-detection-adas-binary-0001	Bounding box: (629, 310) (934, 811), (392, 435) (440, 525)	Bounding box (629, 310) (934, 811), (392, 435) (440, 525)

Test image #10

Data source: Pascal VOC

Image resolution: 500 x 375

Bounding boxes (upper left and bottom right corners):
AEROPLANE (127, 62), (251, 443)

Model	Python (latency mode, implementation)	Python (throughput mode, implementation)
yolo-v2-ava-0001	Bounding box: AEROPLANE (127, 62), (251, 443)	Bounding box: AEROPLANE (127, 62), (251, 443)
yolo-v2-ava-sparse-35-0001	Bounding box: AEROPLANE (129, 19), (258, 410)	Bounding box: AEROPLANE (129, 19), (258, 410)
yolo-v2-ava-sparse-70-0001	Bounding box: AEROPLANE (100, 66), (222, 450)	Bounding box: AEROPLANE (100, 66), (222, 450)
yolo-v2-tiny-ava-0001	Bounding box: AEROPLANE (96, 51), (223, 464)	Bounding box: AEROPLANE (96, 51), (223, 464)
yolo-v2-tiny-ava-sparse-30-0001	Bounding box: AEROPLANE (118, -6), (267, 440)	Bounding box: AEROPLANE (118, -6), (267, 440)
yolo-v2-tiny-ava-sparse-60-0001	Bounding box: AEROPLANE (94, 42), (225, 473)	Bounding box: AEROPLANE (94, 42), (225, 473)

Action detection and recognition

Test image #1

Data source: sample-videos

Image resolution: 1920 x 1080

Bounding boxes (upper left and bottom right corners) and actions:
sitting (1157,517) (1407,1057)
sitting (452,495) (627,874)
sitting (201,555) (469,1084)
raising hand (874,444) (1052,849)

Model	Python (latency mode, implementation)	Python (throughput mode, implementation)
person-detection-action-recognition-0006	Bounding box and action: sitting (1157,517) (1407,1057) sitting (452,495) (627,874) sitting (201,555) (469,1084) raising hand (874,444) (1052,849)	Bounding box and action: sitting (1157,517) (1407,1057) sitting (452,495) (627,874) sitting (201,555) (469,1084) raising hand (874,444) (1052,849)
person-detection-action-recognition-0005	Bounding box and action: sitting (1160,528) (1409,1082) sitting (202,569) (455,1079) standing (453,495) (624,869) raising hand (836,404) (1048,862)	Bounding box and action: sitting (1160,528) (1409,1082) sitting (202,569) (455,1079) standing (453,495) (624,869) raising hand (836,404) (1048,862)
person-detection-raisinghand-recognition-0001	Bounding box and action: sitting (1160,528) (1409,1082) sitting (202,569) (455,1079) sitting (453,495) (624,869) other (836,404) (1048,862)	Bounding box and action: sitting (1160,528) (1409,1082) sitting (202,569) (455,1079) sitting (453,495) (624,869) other (836,404) (1048,862)

Test image #2

Data source: Internet

Image resolution: 1920 x 1080

Bounding boxes (upper left and bottom right corners) and actions:
standing (186,15) (276,224)

Model	Python (latency mode, implementation)	Python (throughput mode, implementation)
person-detection-action-recognition-teacher-0002	Bounding box and action: standing (286,84) (357,283) standing (0,81) (101,281) standing (186,15) (276,224)	Bounding box and action: standing (286,84) (357,283) standing (0,81) (101,281) standing (186,15) (276,224)

Object recognition

Test image #1

Data source: GitHub

Image resolution: 62 x 62

Model	Python (latency mode, implementation)	Python (throughput mode, implementation)
age-gender-recognition-retail-0013	Female, 25.19	Female, 25.19

Test image #2

Data source: GitHub

Image resolution: 62 x 62

Model	Python (latency mode, implementation)	Python (throughput mode, implementation)
age-gender-recognition-retail-0013	Male, 43.43	Male, 43.43

Test image #3

Data source: GitHub

Image resolution: 62 x 62

Model	Python (latency mode, implementation)	Python (throughput mode, implementation)
age-gender-recognition-retail-0013	Male, 28.49	Male, 28.49

Test image #4

Data source: VGGFace2

Image resolution: 48 x 48

Face landmarks:
EYE (17, 18),
EYE (35, 21),
NOSE (24, 27),
LIP CORNER (15, 34),
LIP CORNER (28, 36)

Model	Python (latency mode, implementation)	Python (throughput mode, implementation)
landmarks-regression-retail-0009	Face landmarks: EYE (17, 18), EYE (35, 21), NOSE (24, 27), LIP CORNER (15, 34), LIP CORNER (28, 36)	Face landmarks: EYE (17, 18), EYE (35, 21), NOSE (24, 27), LIP CORNER (15, 34), LIP CORNER (28, 36)

Test image #5

Data source: Internet

Image resolution: 60 x 60

Face landmarks:
LEFT EYE (17, 22), (9, 22),
RIGHT EYE (30, 21), (39, 20),
NOSE (21, 33), (23, 37), (17, 35), (30, 34),
MOUTH (17, 44), (34, 42), (23, 41), (24, 48),
LEFT EYEBROW (6, 17), (11, 15), (18, 17),
RIGHT EYEBROW (27, 15), (35, 12), (43, 14),
FACE CONTOUR (5, 22), (5, 28), (6, 33), (8, 38), (10, 43), (12, 48), (16, 52), (20, 56), (25, 57), (33, 56), (39, 53), (44, 48), (49, 43), (51, 38), (52, 31), (53, 25), (53, 18)

Model	Python (latency mode, implementation)	Python (throughput mode, implementation)
facial-landmarks-35-adas-0002	Face landmarks: LEFT EYE (17, 22), (9, 22), RIGHT EYE (30, 21), (39, 20), NOSE (21, 33), (23, 37), (17, 35), (30, 34), MOUTH (17, 44), (34, 42), (23, 41), (24, 48), LEFT EYEBROW (6, 17), (11, 15), (18, 17), RIGHT EYEBROW (27, 15), (35, 12), (43, 14), FACE CONTOUR (5, 22), (5, 28), (6, 33), (8, 38), (10, 43), (12, 48), (16, 52), (20, 56), (25, 57), (33, 56), (39, 53), (44, 48), (49, 43), (51, 38), (52, 31), (53, 25), (53, 18)	Face landmarks: LEFT EYE (17, 22), (9, 22), RIGHT EYE (30, 21), (39, 20), NOSE (21, 33), (23, 37), (17, 35), (30, 34), MOUTH (17, 44), (34, 42), (23, 41), (24, 48), LEFT EYEBROW (6, 17), (11, 15), (18, 17), RIGHT EYEBROW (27, 15), (35, 12), (43, 14), FACE CONTOUR (5, 22), (5, 28), (6, 33), (8, 38), (10, 43), (12, 48), (16, 52), (20, 56), (25, 57), (33, 56), (39, 53), (44, 48), (49, 43), (51, 38), (52, 31), (53, 25), (53, 18)

Test image #6

Data source: Cityscapes

Image resolution: 80 x 160

Model	Python (latency mode, implementation)	Python (throughput mode, implementation)
person-attributes-recognition-crossroad-0230

Test image #7

Data source: Cityscapes

Image resolution: 80 x 160

Model	Python (latency mode, implementation)	Python (throughput mode, implementation)
person-attributes-recognition-crossroad-0230

Test image #8

Data source: BKHD

Image resolution: 60 x 60

Model	Python (latency mode, implementation)	Python (throughput mode, implementation)
head-pose-estimation-adas-0001

Test image #9

Data source: BKHD

Image resolution: 60 x 60

Model	Python (latency mode, implementation)	Python (throughput mode, implementation)
gaze-estimation-adas-0002

Test image #10

Data source: GitHub

Image resolution: 24 x 94

Model	Python (latency mode, implementation)	Python (throughput mode, implementation)
license-plate-recognition-barrier-0001	<Beijing>FA9512	<Beijing>FA9512

Image processing

Test image #1

Data source: GitHub

Image resolution: 720 x 480

Processed images are identical.

Model	Python (latency mode, implementation)	Python (throughput mode, implementation)
single-image-super-resolution-1032
single-image-super-resolution-1033

Pose recognition

Test image #1

Data source: MS COCO

Image resolution: 640 x 425

Processed images are identical.

Model	Python (latency mode, implementation)	Python (throughput mode, implementation)
human-pose-estimation-0001

Semantic segmentation

Test image #1

Data source: Cityscapes

Image resolution: 2048 x 1024

Segmented images are identical.

Model	Python (latency mode, implementation)	Python (throughput mode, implementation)
semantic-segmentation-adas-0001

Color map:

Test image #2

Data source: GitHub

Image resolution: 640 x 365

Segmented images are identical.

Model	Python (latency mode, implementation)	Python (throughput mode, implementation)
road-segmentation-adas-0001

Color map:

Test image #3

Data source: CamVid

Image resolution: 960 x 720

Segmented images are identical.

Model	Python (latency mode, implementation)	Python (throughput mode, implementation)
unet-camvid-onnx-0001
icnet-camvid-ava-0001
icnet-camvid-ava-sparse-30-0001
icnet-camvid-ava-sparse-60-0001

Color map:

High-level description

Test image #1

Data source: LFW

Image resolution: 250 x 250

Model	Python (latency mode, implementation)	Python (throughput mode, implementation)
face-reidentification-retail-0095	-0.1658423 -0.5230426 -1.4679441 0.0983598 ... 0.8537527 0.8713884 -0.8769233 0.6840097 Full tensor	-0.1658423 -0.5230426 -1.4679441 0.0983598 ... 0.8537527 0.8713884 -0.8769233 0.6840097 Full tensor

Test image #2

Data source: GitHub

Image resolution: 960 x 720

Model	Python (latency mode, implementation)	Python (throughput mode, implementation)
action-recognition-0001-encoder	0.0794002 0.0583136 0.0020747 0.0903931 ... 0.0785143 0.0922345 0.0033597 0.3115494 Full tensor	0.0794002 0.0583136 0.0020747 0.0903931 ... 0.0785143 0.0922345 0.0033597 0.3115494 Full tensor

Test image #3

Data source: GitHub

Image resolution: 1922 x 1080

Model	Python (latency mode, implementation)	Python (throughput mode, implementation)
driver-action-recognition-adas-0002-encoder	-0.0142664 -0.0064784 -0.0334583 -0.0108943 ... -0.2324419 0.2686763 0.0168234 0.0029897 Full tensor	-0.0142664 -0.0064784 -0.0334583 -0.0108943 ... -0.2324419 0.2686763 0.0168234 0.0029897 Full tensor

Test image #4

Data source: Internet

Image resolution: 1922 x 1080

Model	Python (latency mode, implementation)	Python (throughput mode, implementation)
image-retrieval-0001	0.1158277 -0.0189930 0.0530676 0.0290345 ... 0.2057585 -0.0367919 -0.0067885 -0.0031499 Full tensor	0.1158277 -0.0189930 0.0530676 0.0290345 ... 0.2057585 -0.0367919 -0.0067885 -0.0031499 Full tensor

Test image #5

Data source: Internet

Image resolution: 128 x 256

Model	Python (latency mode, implementation)	Python (throughput mode, implementation)
person-reidentification-retail-0277	-0.5144883 -0.4489283 0.1324019 0.2539501 ... 0.4898967 0.1124130 -0.1284953 0.0117971 Full tensor	Errors in output
person-reidentification-retail-0286	0.2997471 -0.2456339 -0.1295844 -0.2274195 ... 0.2052885 0.1565714 0.2504670 0.2383912 Full tensor	Errors in output
person-reidentification-retail-0287	-0.2050491 -0.4432146 0.6389340 0.2023722 ... -0.6498539 -0.0128914 -0.5972998 -0.3941978 Full tensor	Errors in output
person-reidentification-retail-0288	0.2380135 0.3185425 -0.3636540 0.3864555 ... -0.3466439 -0.3920009 -0.0821614 0.1705070 Full tensor	Errors in output

Action recognition

Tensor #1

Data source: output tensor of the action-recognition-0001-encoder model

0.0794002 0.0583136 0.0020747 0.0903931
0.0154800 0.3712009 0.4007360 0.0830761
...
0.1126685 0.1257046 0.1392988 0.5075323
0.0785143 0.0922345 0.0033597 0.3115494

Model	Python (latency mode, implementation)	Python (throughput mode, implementation)
action-recognition-0001-decoder	9.0227661 tying bow tie 7.5208311 tying tie 4.8729849 sign language interpreting 4.3601480 answering questions 4.2990689 tying knot (not on a tie) 4.0868192 whistling 3.9643712 playing harmonica 3.7044604 stretching arm 3.5711651 strumming guitar 3.5514102 playing clarinet	9.0227661 tying bow tie 7.5208311 tying tie 4.8729849 sign language interpreting 4.3601480 answering questions 4.2990689 tying knot (not on a tie) 4.0868192 whistling 3.9643712 playing harmonica 3.7044604 stretching arm 3.5711651 strumming guitar 3.5514102 playing clarinet

Tensor #2

Data source: output tensor of the driver-action-recognition-adas-0002-encoder model

-0.0142664 -0.0064780 -0.0334583 -0.0108943
-0.0555940 -0.0013968 0.0001638 -0.0007524
...
-0.0093990 -0.0028726 0.0074722 0.0303789
-0.2324419 0.2686763 0.0168234 0.0029897

Model	Python (latency mode, implementation)	Python (throughput mode, implementation)
driver-action-recognition-adas-0002-decoder	4.3797836 texting by right hand 4.1073933 talking on the phone by right hand 1.6492549 drinking 1.2682760 texting by left hand 0.3225771 reaching behind -1.6658649 safe driving -3.3440599 doing hair or making up -4.6270852 operating the radio -5.3927083 talking on the phone by left hand	4.3797836 texting by right hand 4.1073933 talking on the phone by right hand 1.6492549 drinking 1.2682760 texting by left hand 0.3225771 reaching behind -1.6658649 safe driving -3.3440599 doing hair or making up -4.6270852 operating the radio -5.3927083 talking on the phone by left hand

Instance segmentation

Test image #1

Data source: MS COCO

Image resolution: 640 x 480

Input tensor: 480; 640; 1

Model	Python (latency mode, implementation)	Python (throughput mode, implementation)
instance-segmentation-security-0083

Test image #2

Data source: MS COCO

Image resolution: 640 x 640

Input tensor: 480; 480; 1

Model	Python (latency mode, implementation)	Python (throughput mode, implementation)
instance-segmentation-security-0050
instance-segmentation-security-1025

Test image #3

Data source: MS COCO

Image resolution: 640 x 427

Input tensor: 800; 1344; 1

Model	Python (latency mode, implementation)	Python (throughput mode, implementation)
instance-segmentation-security-0010

Color map:

Image classification

Test image #1

Data source: ImageNet

Image resolution: 709 x 510

Model	Python (latency mode, implementation)	Python (throughput mode, implementation)
resnet18-xnor-binary-onnx-0001	6.5452480 Granny Smith 4.1318626 fig 3.5715680 bell pepper 3.1780813 saltshaker, salt shaker 3.1212788 hair slide	6.5452480 Granny Smith 4.1318626 fig 3.5715680 bell pepper 3.1780813 saltshaker, salt shaker 3.1212788 hair slide

Test image #2

Data source: ImageNet

Image resolution: 500 x 500

Model	Python (latency mode, implementation)	Python (throughput mode, implementation)
resnet18-xnor-binary-onnx-0001	9.1701651 junco, snowbird 5.4874449 chickadee 0.4869275 jay 0.3719085 indigo bunting, indigo finch, indigo bird, Passerina cyanea -1.1992515 brambling, Fringilla montifringilla	9.1701651 junco, snowbird 5.4874449 chickadee 0.4869275 jay 0.3719085 indigo bunting, indigo finch, indigo bird, Passerina cyanea -1.1992515 brambling, Fringilla montifringilla

Test image #3

Data source: ImageNet

Image resolution: 333 x 500

Model	Python (latency mode, implementation)	Python (throughput mode, implementation)
resnet18-xnor-binary-onnx-0001	4.7719054 lifeboat 1.7933186 drilling platform, offshore rig 0.1516396 fireboat 0.0121927 amphibian, amphibious vehicle -0.2893910 pirate, pirate ship	4.7719054 lifeboat 1.7933186 drilling platform, offshore rig 0.1516396 fireboat 0.0121927 amphibian, amphibious vehicle -0.2893910 pirate, pirate ship

Files

validation_results_openvino_intel_models.md

Latest commit

History

validation_results_openvino_intel_models.md

File metadata and controls

Validation results for the models inferring using Intel® Distribution of OpenVINO™ Toolkit

Object detection

Test image #1

Test image #2

Test image #3

Test image #4

Test image #5

Test image #6

Test image #7

Test image #8

Test image #9

Test image #10

Action detection and recognition

Test image #1

Test image #2

Object recognition

Test image #1

Test image #2

Test image #3

Test image #4

Test image #5

Test image #6

Test image #7

Test image #8

Test image #9

Test image #10

Image processing

Test image #1

Pose recognition

Test image #1

Semantic segmentation

Test image #1

Test image #2

Test image #3

High-level description

Test image #1

Test image #2

Test image #3

Test image #4

Test image #5

Action recognition

Tensor #1

Tensor #2

Instance segmentation

Test image #1

Test image #2

Test image #3

Image classification

Test image #1

Test image #2

Test image #3