Downstream Semantic Segmentation Model for Low-Level Surface Crack Detection

<div>Some results from our downstream model that is trained by <svg height="12.2872pt" id="M73" style="vertical-align:-3.27288pt" version="1.1" viewbox="-0.0498162 -9.01432 24.6244 12.2872" width="24.6244pt" xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink"><g transform="matrix(.013,0,0,-0.013,0,0)"><path d="M608 643C551 682 486 687 443 687C299 687 188 608 188 504C188 361 341 312 465 312C483 312 501 313 518 315C454 195 366 93 314 59C255 82 201 109 141 109C73 109 33 83 33 52C33 13 76 -9 159 -9C218 -9 267 -1 314 15C354 0 402 -15 478 -15C575 -15 671 36 759 141L740 155C657 63 571 16 475 16C448 16 411 26 372 39C458 74 565 199 627 333C806 381 931 499 931 587C931 634 908 669 848 669C767 669 671 582 620 506C591 462 569 411 534 347C503 343 490 342 471 342C330 342 229 399 229 500C229 603 346 657 451 657C486 657 540 657 596 624L608 643ZM890 588C890 519 822 457 747 414C712 395 678 378 643 369C683 484 760 632 841 632C873 632 890 615 890 588ZM266 35C235 24 200 21 161 21C103 21 74 32 74 54C74 75 122 79 141 79C172 79 209 69 266 36V35Z"></path></g><g transform="matrix(.0091,0,0,-0.0091,12.298,3.132)"><path d="M152 404V712C115 698 54 683 7 677V654C71 648 73 642 73 579V24C128 -2 179 -12 220 -12C353 -12 471 92 471 238C471 357 381 449 274 449C262 449 249 446 233 439L152 404ZM152 374C170 384 202 393 230 393C313 393 382 326 382 213C382 97 330 26 246 26C194 26 165 62 158 81C154 91 152 101 152 116V374Z"></path></g><g transform="matrix(.0091,0,0,-0.0091,16.811,3.132)"><path d="M54 437L27 408L31 397H101V103C101 37 94 32 30 26V0H266V25C187 33 180 36 180 110V397H288C299 404 304 428 298 437H180V477C179 562 190 610 203 630C214 647 230 659 256 659C289 659 318 641 337 622C346 612 355 612 364 619C374 627 380 635 383 643C388 655 387 667 378 678C362 697 333 710 299 712C260 707 225 689 189 659C135 613 119 563 112 541S101 490 101 458V437H54Z"></path></g><g transform="matrix(.0091,0,0,-0.0091,19.532,3.132)"><path d="M414 144C384 79 371 75 317 75H135L276 221C367 316 408 376 408 465C408 570 327 635 237 635C179 635 131 609 100 575L42 494L67 471C94 510 138 565 205 565C277 565 321 517 321 435C321 348 258 270 195 195C146 137 88 81 33 26V0H411C423 44 433 88 446 135L414 144Z"></path></g></svg> from random weights. For better visualization, binary images in the second and fourth columns are inverted and greyscale images in the third column are colorized with the jet colormap in matplotlib.</div>

Advances in Multimedia

fig4

Figure 4

Figure 4: Downstream Semantic Segmentation Model for Low-Level Surface Crack Detection 