{"id":177,"date":"2023-09-07T14:23:30","date_gmt":"2023-09-07T12:23:30","guid":{"rendered":"https:\/\/ics.science.upjs.sk\/ano\/?page_id=177"},"modified":"2024-09-20T04:03:12","modified_gmt":"2024-09-20T02:03:12","slug":"domace-zadania-2023","status":"publish","type":"page","link":"https:\/\/ics.science.upjs.sk\/ano\/domace-zadania-2023\/","title":{"rendered":"Dom\u00e1ce zadania 2023"},"content":{"rendered":"<p>Podmienky odovzd\u00e1vania dom\u00e1cich zadan\u00ed:<\/p>\n<ul>\n<li>Pre ka\u017ed\u00fa \u00falohu o\u010dak\u00e1vam zdrojov\u00fd k\u00f3d, vstupn\u00fd obr\u00e1zok\/video a v\u00fdstupn\u00fd obr\u00e1zok\/video (pr\u00edpadne v\u0161etky medzi-obr\u00e1zky) \u2013 d\u00f4le\u017eit\u00e9 je, aby som mohol vidie\u0165 v\u00fdsledok bez nutnosti sp\u00fa\u0161\u0165a\u0165 k\u00f3d.<\/li>\n<li>Rie\u0161enia odovzd\u00e1vate cez Github classroom. Na rie\u0161enie zadan\u00ed bude zhruba 10 dn\u00ed. Zadanie bude zverejnen\u00e9 v piatky pod\u013ea harmonogramu, term\u00edn odovzdania bude v\u017edy pondelok r\u00e1no. Feedback budem d\u00e1va\u0165 v github classroom.<\/li>\n<li>Po\u010das semestra bude zverejnen\u00fdch 5 \u00faloh (2x v okt\u00f3bri, 2x v novembri, 1x v decembri). Ob\u013e\u00faben\u00e9 \u00falohy z minul\u00e9ho roka (dom\u00e1ce zadania, sk\u00fa\u0161ka) sa m\u00f4\u017eu zopakova\u0165. \u00dalohy bud\u00fa zverejnen\u00e9 v piatok dopoludnia, deadline je o 10 dn\u00ed, v pondelok r\u00e1no\n<ul>\n<li>1. \u00faloha, zverejnen\u00e1 20.10. &#8211; deadline 30.10. 8:00<\/li>\n<li>2. \u00faloha, zverejnen\u00e1 27.10. &#8211; deadline 6.11. 8:00<\/li>\n<li>3. \u00faloha, zverejnen\u00e1 10.11. &#8211; deadline 20.11. 8:00<\/li>\n<li>4. \u00faloha, zverejnen\u00e1 24.11. &#8211; deadline 4.12. 8:00<\/li>\n<li>5. \u00faloha, zverejnen\u00e1 8.12. &#8211; deadline 18.12. 8:00<\/li>\n<\/ul>\n<\/li>\n<\/ul>\n<hr \/>\n<h2>1. \u00faloha: Ma\u013eovan\u00e9 kr\u00ed\u017eovky (<a href=\"https:\/\/classroom.github.com\/a\/YGt1ANsp\">LINK na github classroom<\/a>)<\/h2>\n<p><em>Term\u00edn odovzdania: 30.10.2023 8:00<\/em><\/p>\n<p>Vstupn\u00fdm obr\u00e1zkom je fotografia zatia\u013e nevyrie\u0161enej ma\u013eovanej kr\u00ed\u017eovky. Cie\u013eom je identifikova\u0165 mrie\u017eku a \u010d\u00edsla a n\u00e1sledne z toho vytvori\u0165 digit\u00e1lnu verziu kr\u00ed\u017eovky. Na rozpozn\u00e1vanie \u010d\u00edsel existuj\u00fa met\u00f3dy (zalo\u017een\u00e9 na strojovom u\u010den\u00ed, pr\u00edpadne template matching) a tie\u017e na n\u00e1jdenie \u010diar existuj\u00fa sp\u00f4soby (Hough transform\u00e1cia na detekciu \u010diar, LSD algoritmus) &#8211; t\u00fdmto veciam sa budeme venova\u0165 nesk\u00f4r a m\u00f4\u017eeme si na cvi\u010deniach vysk\u00fa\u0161a\u0165 ak\u00fd v\u00fdsledok daj\u00fa na t\u00fdchto obr\u00e1zkoch.<\/p>\n<p>Obr\u00e1zky s\u00fa vyfoten\u00e9 telef\u00f3nom a z\u00e1merne s\u00fa niektor\u00e9 hor\u0161ej kvality (osvetlenie, rozostrenie a pod.). V t\u00fdchto pr\u00edpadoch detekcia \u010diar nemus\u00ed by\u0165 ide\u00e1lna a jednoduch\u0161ie by mohlo by\u0165 aplikova\u0165 t\u00fato detekciu na bin\u00e1rnom obraze.<\/p>\n<p>Va\u0161ou \u00falohou je zo zadan\u00e9ho obr\u00e1zka vytvori\u0165 bin\u00e1rny obr\u00e1zok, kde v najlep\u0161om pr\u00edpade bud\u00fa bielou farbou zobrazen\u00e9 \u010diary a \u010diernou v\u0161etko ostatn\u00e9. \u00daloha je to n\u00e1ro\u010dn\u00e1, pok\u00faste sa o \u010do najlep\u0161\u00ed v\u00fdsledok. Postup si m\u00f4\u017eete vybra\u0165 \u013eubovo\u013en\u00fd. Nieko\u013eko tipov:<\/p>\n<ul>\n<li>zmena rozl\u00ed\u0161enia &#8211; \u010dasto je vhodnej\u0161ie pracova\u0165 s obr\u00e1zkom v men\u0161om rozl\u00ed\u0161en\u00ed. M\u00f4\u017eete si obr\u00e1zky zmen\u0161i\u0165 manu\u00e1lne alebo pomocou pr\u00edkazu resize v opencv.<\/li>\n<li>preprocessing &#8211; \u00faprava jasu, vyhladenie histogramu, aplik\u00e1cia nejak\u00e9ho <a href=\"https:\/\/en.wikipedia.org\/wiki\/Kernel_(image_processing)\">filtra<\/a> (pozrite sharpen)<\/li>\n<li>zmena na bin\u00e1rny obr\u00e1zok &#8211; thresholding, canny detektor<\/li>\n<li>postprocessing &#8211; dilat\u00e1cia, er\u00f3zia (porozm\u00fd\u0161\u013eajte nad tvarom \u0161trukt\u00farneho elementu)<\/li>\n<\/ul>\n<p>M\u00f4\u017eete si zvoli\u0165 \u013eubovo\u013en\u00fa strat\u00e9giu, napr.:<\/p>\n<ul>\n<li>vybra\u0165 si najjednoduch\u0161\u00ed obr\u00e1zok. Vysk\u00fa\u0161a\u0165 kombin\u00e1ciu met\u00f3d, ktor\u00e9 daj\u00fa najlep\u0161\u00ed v\u00fdsledok. N\u00e1sledne to otestova\u0165 aj na zlo\u017eitej\u013e\u0161\u00edch vstupoch. Pr\u00edpadne upravi\u0165 postup.<\/li>\n<li>vybra\u0165 si niektor\u00fd scen\u00e1r (napr. zl\u00e9 osvetlenie) a \u0161peci\u00e1lne sa venova\u0165 tomuto pr\u00edpadu.<\/li>\n<li>od za\u010diatku sk\u00fa\u0161a\u0165 postup, ktor\u00fd bude fungova\u0165 na viacer\u00fdch pr\u00edpadoch<\/li>\n<\/ul>\n<p>Vo v\u00fdsledku ozna\u010dte, s ktor\u00fdm obr\u00e1zkom ste pracovali najviac. V\u00fdsledn\u00e1 met\u00f3da nech je spusten\u00e1 na v\u0161etk\u00fdch obr\u00e1zkoch (aj ke\u010f to dosiahne nedobr\u00fd v\u00fdsledok). V\u00fdstupn\u00e9 obr\u00e1zky odovzdajte spolu s k\u00f3dom (aby som k\u00f3d nemusel sp\u00fa\u0161\u0165a\u0165 &#8211; iba si ho pozrie\u0165).<\/p>\n<p>O\u010dak\u00e1vam, \u017ee ka\u017ed\u00fd vypracuje zadanie samostatne. Samozrejme ist\u00e1 miera spolupr\u00e1ce m\u00f4\u017ee by\u0165 v\u00edtan\u00e1, ozna\u010dte to pros\u00edm v rie\u0161en\u00ed ak ste spolupracovali s niek\u00fdm in\u00fdm.<\/p>\n<h2>2. \u00faloha: RTG detekcia zlomeniny (<a href=\"https:\/\/classroom.github.com\/a\/nc4HzVlD\">LINK na github classroom<\/a>)<\/h2>\n<p><em>Term\u00edn odovzdania: 6.11.2023 8:00<\/em><\/p>\n<p>Z\u013eahka si pre\u010d\u00edtajte nasledovn\u00fd vedeck\u00fd pr\u00edspevok:<\/p>\n<p><em>Basha, Cmak Zeelan, et al. &#8222;Enhanced computer aided bone fracture detection employing X-ray images by Harris Corner technique.&#8220; 2020 Fourth International Conference on Computing Methodologies and Communication (ICCMC). IEEE, 2020. (<\/em><a href=\"https:\/\/ics.science.upjs.sk\/ano\/wp-content\/uploads\/sites\/17\/2023\/10\/basha2020enhanced.pdf\">PDF<\/a>)<\/p>\n<p>Implementujte pop\u00edsan\u00fd postup:<\/p>\n<ul>\n<li>Preprocessing s pou\u017eit\u00edm Gaussian filtering. Rozostrenie obr\u00e1zka.<\/li>\n<li>Segment\u00e1cia pomocou Cannyho hranov\u00e9ho detektora.<\/li>\n<li>Feature extraction s pou\u017eit\u00edm Harris rohov\u00e9ho detektora.<\/li>\n<li>Klasifik\u00e1cia, \u010di je na obr\u00e1zku zlomenina (toto nemus\u00ed by\u0165 implementovan\u00e9 v k\u00f3de &#8211; sta\u010d\u00ed slovn\u00fd popis).<\/li>\n<\/ul>\n<p>V Github classroom n\u00e1jdete nejak\u00e9 vstupn\u00e9 obr\u00e1zky. Vybral som 3, na ktor\u00fdch je dobre vidie\u0165 zlomenina. Plus dve bez zlomeniny. M\u00f4\u017eete si doplni\u0165 vlastn\u00e9 vstupn\u00e9 obr\u00e1zky. <a href=\"https:\/\/figshare.com\/articles\/dataset\/The_dataset\/22363012\">Dataset FracAtlas<\/a> je dostupn\u00fd na stiahnutie (322MB). Sn\u00edmky so zlomeninou s\u00fa v osobitnom prie\u010dinku.<\/p>\n<p>Moje postrehy po pre\u010d\u00edtan\u00ed \u010dl\u00e1nku:<\/p>\n<ul>\n<li>Matematiku za t\u00fdm nemus\u00edte \u010d\u00edta\u0165, je to vysvetlenie t\u00fdch met\u00f3d vo v\u0161eobecnosti. Nie ich pr\u00edstup.<\/li>\n<li>Pri v\u0161etk\u00fdch pou\u017eit\u00fdch met\u00f3dach ch\u00fdbaj\u00fa detaily &#8211; napr. thresholdy na Cannyho detektor. To je \u00faloha pre v\u00e1s.<\/li>\n<li>Nie je mi \u00faplne jasn\u00e9 ako robili t\u00fa samotn\u00fa klasifik\u00e1ciu a ke\u010f som si pozeral podobn\u00e9 \u010dl\u00e1nky, tak m\u00e1m pochybnos\u0165 o v\u00fdsledku tohto pr\u00edstupu.<\/li>\n<\/ul>\n<p>Sk\u00faste sa pozrie\u0165 na v\u00fdstup detekcie rohov pri zlomenin\u00e1ch aj pri zdrav\u00fdch kostiach. K samotn\u00e9mu k\u00f3du prilo\u017ete aj <strong>slovn\u00fd popis<\/strong> ako by ste robili rozhodnutie (klasifik\u00e1ciu).<\/p>\n<p>Rie\u0161enie nech obsahuje:<\/p>\n<ul>\n<li>k\u00f3d (python alebo jupyter notebook)<\/li>\n<li>obr\u00e1zky, ktor\u00e9 ste pou\u017eili &#8211; ak pou\u017eijete ve\u013ea obr\u00e1zkov, nemus\u00edte priklada\u0165 v\u0161etky. Ale prie\u010dinok input by mohol obsahova\u0165 viac ako 5 obr\u00e1zkov, ktor\u00e9 som tam dal ja.<\/li>\n<li>v\u00fdstupn\u00e9 obr\u00e1zky &#8211; aby som nemusel sp\u00fa\u0161\u0165a\u0165 k\u00f3d a videl v\u00fdstup. Nech tam je aspo\u0148 jeden pr\u00edklad zdravej kosti a aspo\u0148 jedna zlomenina. M\u00f4\u017ee by\u0165 priamo v notebooku alebo v\u00fdstupn\u00fd obr\u00e1zok v nejakom prie\u010dinku.<\/li>\n<li>slovn\u00fd popis ako by ste robili klasifik\u00e1ciu (m\u00f4\u017ee by\u0165 priamo pri k\u00f3de, alebo ako samostatn\u00fd textov\u00fd s\u00fabor)<\/li>\n<\/ul>\n<h2>3. \u00faloha: MHz \u2013 Tomoskopia (<a href=\"https:\/\/classroom.github.com\/a\/VRgfw62N\">LINK na github classroom<\/a>)<\/h2>\n<p><em>Term\u00edn odovzdania: 20.11.2023 8:00<\/em><\/p>\n<p>Toto zadanie s\u00favis\u00ed s bakal\u00e1rskou pr\u00e1cou D. Mo\u0161ka. Tu je jeho intro k \u00falohe:<br \/>\n<em>Obr\u00e1zky sme z\u00edskali na experimente MHz Tomoskopie (r\u00f6ntgenovsk\u00e9 zobrazovanie) v Hamburgu. Pou\u017eili sme optick\u00fd syst\u00e9m z germ\u00e1niov\u00fdch kry\u0161t\u00e1lov v setupe. tzv Braggovho zv\u00e4\u010d\u0161ov\u00e1ka. Obr\u00e1zok sa formuje tak, \u017ee X-ray l\u00fa\u010d prech\u00e1dza cez vzorku a n\u00e1sledne t\u00fdmto optick\u00fdm syst\u00e9mom, ktor\u00fd n\u00e1m zv\u00e4\u010d\u0161\u00ed objekt na detektore a\u017e cca. 200x (vodn\u00e1 guli\u010dka na obr\u00e1zkoch m\u00e1 cca 100 \u03bcm). Kv\u00f4li charakteru l\u00fa\u010da, ktor\u00fd funguje na b\u00e1ze vo\u013en\u00fdch elektr\u00f3nov, sa men\u00ed ilumina\u010dn\u00fd profil a intenzita. Spolu s r\u00f4znymi prachov\u00fdmi \u010dasticami, \u0161umu zo samotn\u00e9ho detektora a nie v\u00e1kuov\u00e9ho prostredia sa vytv\u00e1ra na obr\u00e1zku \u0161um a tak zvidite\u013eni\u0165 kru\u017enice (difrak\u010dn\u00e9 kr\u00fa\u017eky = fringes), aby sa mohli \u010falej pou\u017ei\u0165 rekon\u0161truk\u010dn\u00e9 algoritmy.<\/em><\/p>\n<p><em>Pre m\u0148a osobne je zauj\u00edmav\u00e9 odstr\u00e1nenie \u0161umu. Podarilo sa mi z datasetu z\u00edska\u0165 \u0161umy, ktor\u00e9 s\u00fa na obr\u00e1zkoch bez vzorky (pre zauj\u00edmavos\u0165 posielam obr\u00e1zok)<\/em><\/p>\n<p>V github repozit\u00e1ri n\u00e1jdete 19 obr\u00e1zkov, na ktor\u00fdch m\u00f4\u017eete vysk\u00fa\u0161a\u0165 svoje rie\u0161enie. Okrem toho sa tam nach\u00e1dza obr\u00e1zok noise.png s uk\u00e1\u017ekou \u0161umu.<\/p>\n<p>Vzh\u013eadom na dostupnos\u0165 mno\u017estva obr\u00e1zkov a povahu \u00falohy (r\u00fdchlos\u0165 v\u00fdpo\u010dtu) vyzer\u00e1, \u017ee pou\u017eitie neur\u00f3nov\u00fdch siet\u00ed, resp. strojov\u00e9ho u\u010denia je vhodn\u00e9 na dan\u00fd probl\u00e9m. Napriek tomu by bolo fajn zisti\u0165, \u010do je mo\u017en\u00e9 dosiahn\u00fa\u0165 s pou\u017eit\u00edm met\u00f3d po\u010d\u00edta\u010dov\u00e9ho videnia.<\/p>\n<p>Hlavnou \u00falohou je pok\u00fasi\u0165 sa odstr\u00e1ni\u0165 alebo potla\u010di\u0165 \u0161um v obr\u00e1zku. Ak\u00e9ko\u013evek rie\u0161enie je v\u00edtan\u00e9, odo\u0161lite aj rie\u0161enia, ktor\u00e9 ned\u00e1vaj\u00fa dobr\u00fd v\u00fdsledok. Ide\u00e1lne pripojte aj nejak\u00fd koment\u00e1r &#8211; kr\u00e1tky odstavec, kde zhrniete svoje pozorovania.<\/p>\n<p>Zop\u00e1r tipov:<\/p>\n<ul>\n<li>ak neviete ako za\u010da\u0165, sk\u00faste nejak\u00fa konvol\u00faciu. M\u00f4\u017eete vysk\u00fa\u0161a\u0165 r\u00f4zne konvolu\u010dn\u00e9 masky.<\/li>\n<li>mo\u017eno sa d\u00e1 pozrie\u0165 na fourierov\u00fa transform\u00e1ciu, pr\u00edpadne ju upravi\u0165, aby nebolo nutn\u00e9 manu\u00e1lne klika\u0165 a potl\u00e1\u010da\u0165 \u0161um<\/li>\n<li>m\u00f4\u017eete sk\u00fasi\u0165 poh\u013eada\u0165 nejak\u00fd in\u00fd pr\u00edstup a vysk\u00fa\u0161a\u0165 (pridajte pros\u00edm k rie\u0161eniu aj odkazy na pou\u017eit\u00fa literat\u00faru)<\/li>\n<li>ak chcete, m\u00f4\u017eete vysk\u00fa\u0161a\u0165 detekciu kru\u017en\u00edc (hough transform\u00e1cia) pred a po aplikovan\u00ed nejakej met\u00f3dy na potla\u010denie \u0161umu<\/li>\n<\/ul>\n<p>Vyberte si \u013eubovo\u013ene sp\u00f4sob rie\u0161enia tejto \u00falohy. Pestros\u0165 zvolen\u00fdch pr\u00edstupov je v\u00edtan\u00e1. Nebojte sa experimentova\u0165.<\/p>\n<h2>4. \u00faloha: Chamele\u00f3n (<a href=\"https:\/\/classroom.github.com\/a\/5BnaAnTk\">LINK na github classroom<\/a>)<\/h2>\n<p><em>Term\u00edn odovzdania: 4.12.2023 8:00<\/em><\/p>\n<p>N\u00e1jdite si obr\u00e1zok chamele\u00f3na. Aplikujte met\u00f3du GrabCut na segment\u00e1ciu &#8211; oddelenie chamele\u00f3na od pozadia. Odpor\u00fa\u010dam pozrie\u0165 <a href=\"https:\/\/ics.science.upjs.sk\/ano\/wp-content\/uploads\/sites\/17\/2023\/11\/Segmentacia-Grabcut.pdf\">\u0161tudijn\u00fd text<\/a>, <a href=\"https:\/\/opencv24-python-tutorials.readthedocs.io\/en\/latest\/py_tutorials\/py_imgproc\/py_grabcut\/py_grabcut.html\">n\u00e1vod<\/a> a <a href=\"https:\/\/github.com\/jasonyzhang\/interactive_grabcut\/tree\/main\">uk\u00e1\u017eku k\u00f3du<\/a>.<\/p>\n<p>Vymyslite sp\u00f4sob ako zmeni\u0165 farbu chamele\u00f3na. Nech\u00e1vam to na va\u0161ej kreativite. Moja predstava je, \u017ee vyr\u00e1tate nejak\u00fa charakteristick\u00fa hodnotu farby pozadia a na z\u00e1klade tejto hodnoty uprav\u00edte chamele\u00f3na. Napr. zjednodu\u0161ene povedan\u00e9, ak je pozadie \u010derven\u00e9 a chamele\u00f3n zelen\u00fd, tak sa p\u00f4vodna farba pixelov patriacich chamele\u00f3novi modifikuje smerom k \u010dervenej farbe (v\u00fdsledn\u00e1 hodnota je nejak\u00e1 kombin\u00e1cia p\u00f4vodnej hodnoty a hodnoty vyr\u00e1tanej z pozadia).<\/p>\n<h2>5. \u00faloha: Vanishing point (<a href=\"https:\/\/classroom.github.com\/a\/UGtDsk_l\">LINK na github classroom<\/a>)<\/h2>\n<p><em>Term\u00edn odovzdania: 18.12.2023 8:00<\/em><\/p>\n<p>Implementujte algoritmus pop\u00edsan\u00fd \u010dl\u00e1nku (<em>Ebrahimpour, R., Rasoolinezhad, R., Hajiabolhasani, Z. and Ebrahimi, M., 2012. Vanishing point detection in corridors: using Hough transform and K-means clustering. IET computer vision, 6(1), pp.40-51.<\/em>).<\/p>\n<p>\u010cl\u00e1nok na stiahnutie: <a href=\"https:\/\/media.proquest.com\/media\/hms\/ORIG\/1\/2ddwB?cit%3Aauth=Ebrahimpour%2C+R%3BRasoolinezhad%2C+R%3BHajiabolhasani%2C+Z%3BEbrahimi%2C+M&amp;cit%3Atitle=Vanishing+point+detection+in+corridors%3A+using+Hough+transform+and+K-means+clustering&amp;cit%3Apub=IET+Computer+Vision&amp;cit%3Avol=6&amp;cit%3Aiss=1&amp;cit%3Apg=40&amp;cit%3Adate=Jan+2012&amp;ic=true&amp;cit%3Aprod=ProQuest+Central&amp;_a=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%2BgIBToIDA1dlYooDHENJRDoyMDIzMTIwODEwMjQyNzg2OToyMzQyOTE%3D&amp;_s=hrwXT2kS5PpFcVD%2Bm9K7nYEM0Bk%3D\">tu<\/a> alebo <a href=\"https:\/\/www.researchgate.net\/profile\/Zeinab-Hajiabolhasani\/publication\/260584378_Vanishing_point_detection_in_corridors_Using_Hough_transform_and_K-means_clustering\/links\/53e423e70cf2fb748710b0b9\/Vanishing-point-detection-in-corridors-Using-Hough-transform-and-K-means-clustering.pdf\">tu<\/a> (ak by bol probl\u00e9m so stiahnut\u00edm, nap\u00ed\u0161te mi, prepo\u0161lem pdf).<\/p>\n<p>Konkr\u00e9tne kroky:<\/p>\n<ol>\n<li>N\u00e1jdite si obr\u00e1zok chodby. M\u00f4\u017eete prisp\u00f4sobi\u0165 rozl\u00ed\u0161enie. M\u00f4\u017eete pou\u017ei\u0165 aj vlastn\u00fa fotku nejakej chodby. Posta\u010duje jeden obr\u00e1zok, ide\u00e1lne vysk\u00fa\u0161a\u0165 aspo\u0148 na 2-3 obr\u00e1zkoch.<\/li>\n<li>Obr\u00e1zok konvertujte na \u0161edot\u00f3nov\u00fd<\/li>\n<li>Ekvalizujte histogram<\/li>\n<li>Aplikujte 45\u00b0 masku (je v \u010dl\u00e1nku &#8211; ide o maticu 3&#215;3 kde na diagon\u00e1le s\u00fa 2 a na ostatn\u00fdch miestach -1), [[-1,-1,2],[-1,2,-1],[2,-1,-1]]<\/li>\n<li>Aplikujte Cannyho detektor na detekciu hr\u00e1n<\/li>\n<li>Pou\u017eite Hough transform\u00e1ciu na z\u00edskanie \u010diar. Odpor\u00fa\u010dam <a href=\"https:\/\/vovkos.github.io\/doxyrest-showcase\/opencv\/sphinxdoc\/page_tutorial_py_houghlines.html\">HoughLinesP<\/a>, kde je v\u00fdstupom zoznam \u010diar, ktor\u00e9 s\u00fa tvoren\u00e9 \u0161tvoricou \u010d\u00edsel (x a y s\u00faradnice za\u010diato\u010dn\u00e9ho a koncov\u00e9ho bodu)<\/li>\n<li>Aplikujte K-means (k=4) na mno\u017eine bodov &#8211; koncov\u00fdch bodov detegovan\u00fdch \u010diar. V\u00fdsledn\u00e9 centroidy pre klastre vykreslite do p\u00f4vodn\u00e9ho obr\u00e1zka.<\/li>\n<li>Aplikujte K-means (k=1) na mno\u017eine 4 centroidov. Zakreslite v\u00fdsledn\u00fd centroid &#8211; v ide\u00e1lnom pr\u00edpade by to mal by\u0165 vanishing point. Vo v\u00fdslednom obr\u00e1zku nech s\u00fa iba poz\u00edcie centroidov. Av\u0161ak vykreslite osobitne aj \u010diastkov\u00e9 rie\u0161enia vr\u00e1tane v\u00fdstupu po detekcii \u010diar.<\/li>\n<\/ol>\n","protected":false},"excerpt":{"rendered":"<p>Podmienky odovzd\u00e1vania dom\u00e1cich zadan\u00ed: Pre ka\u017ed\u00fa \u00falohu o\u010dak\u00e1vam zdrojov\u00fd k\u00f3d, vstupn\u00fd obr\u00e1zok\/video a v\u00fdstupn\u00fd obr\u00e1zok\/video (pr\u00edpadne v\u0161etky medzi-obr\u00e1zky) \u2013 d\u00f4le\u017eit\u00e9 je, aby som mohol vidie\u0165 v\u00fdsledok bez nutnosti sp\u00fa\u0161\u0165a\u0165 k\u00f3d.&hellip;<\/p>\n","protected":false},"author":22,"featured_media":0,"parent":0,"menu_order":0,"comment_status":"closed","ping_status":"closed","template":"","meta":{"footnotes":""},"class_list":["post-177","page","type-page","status-publish","hentry"],"_links":{"self":[{"href":"https:\/\/ics.science.upjs.sk\/ano\/wp-json\/wp\/v2\/pages\/177","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/ics.science.upjs.sk\/ano\/wp-json\/wp\/v2\/pages"}],"about":[{"href":"https:\/\/ics.science.upjs.sk\/ano\/wp-json\/wp\/v2\/types\/page"}],"author":[{"embeddable":true,"href":"https:\/\/ics.science.upjs.sk\/ano\/wp-json\/wp\/v2\/users\/22"}],"replies":[{"embeddable":true,"href":"https:\/\/ics.science.upjs.sk\/ano\/wp-json\/wp\/v2\/comments?post=177"}],"version-history":[{"count":23,"href":"https:\/\/ics.science.upjs.sk\/ano\/wp-json\/wp\/v2\/pages\/177\/revisions"}],"predecessor-version":[{"id":317,"href":"https:\/\/ics.science.upjs.sk\/ano\/wp-json\/wp\/v2\/pages\/177\/revisions\/317"}],"wp:attachment":[{"href":"https:\/\/ics.science.upjs.sk\/ano\/wp-json\/wp\/v2\/media?parent=177"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}