Zviri Mukati[Viga][Ratidza]
- 1. CelebFaces Attributes Dataset
- 2. DOTA
- 3. Google Facial Expression comparison dataset
- 4. Visual Genome
- 5. LibriSpeech
- 6. Nzvimbo dzeMaguta
- 7. Kinetics Dataset
- 8. CelebAMask-HQ
- 9. Penn Treebank
- 10. VoxCeleb
- 11. SIXray
- 12. Njodzi dzeUS
- 13. Ocular Disease Kuzivikanwa
- 14. Hosha Yemwoyo
- 15. CLEVR
- 16. Universal Dependencies
- 17. KITTI – 360
- 18. MOT(Multiple Object Tracking)
- 19. PASCAL 3D+
- 20. Facial Deformable Models yeMhuka
- 21. MPII Human Post Dataset
- 22. UCF101
- 23. Audioset
- 24. Stanford Natural Language Inference
- 25. Mhinduro yemibvunzo yekuona
- mhedziso
Mazuvano, vazhinji vedu takatarisana nekugadzira muchina kudzidza uye AI modhi uye kugadzirisa nyaya uchishandisa azvino dataset. Asi chekutanga, isu tinofanirwa kutsanangura dataset, kukosha kwayo, uye basa rayo mukugadzira yakasimba AI neML mhinduro.
Nhasi, tine huwandu hweakavhurika-sosi dhataseti ekuti tiite tsvagiridzo kana kugadzira maapplication ekugadzirisa nyaya dzepasirese muzvikamu zvakasiyana.
Nekudaro, kushomeka kwemaseti emhando yepamusoro dhatabheti ndiko kunonetsa. Data yakakwira zvakanyanya uye icharamba ichiwedzera nekukurumidza mune ramangwana.
Mune ino positi, isu tinovhara mahara anowanikwa dhataseti aunogona kushandisa kugadzira yako inotevera AI chirongwa.
1. CelebFaces Attributes Dataset
CelebFaces Attributes Dataset (CelebA) ine anopfuura mazana maviri emifananidzo ane mukurumbira uye makumi mana zvirevo zvemufananidzo wega wega, zvichiita kuti ive nzvimbo yakanaka yekutanga mapurojekiti akadai. kuzivikanwa kwechiso, kuonekwa kwechiso, kucherechedzwa (kana chikamu chechiso) kugarisana, uye kugadzirisa kwechiso & synthesis. Uyezve, mapikicha ari muunganidzwa uyu ane akasiyana siyana ezvinzvimbo uye nekumashure clutter.
2. Dota
DOTA (Dataset ye Object Detection muAerial Photos) inzvimbo huru yedataset yekuona zvinhu iyo inosanganisira gumi neshanu zvikamu zvakajairika (semuenzaniso, ngarava, ndege, mota, nezvimwewo), 15 mifananidzo yekudzidziswa, uye 1411 mifananidzo yekusimbisa.
3. Google Facial Expression yekuenzanisa dataset
Iyo Google facial expression yekuenzanisa dataset ine anosvika mazana mashanu ezviuru emifananidzo katatu, kusanganisira 500,000 mapikicha echiso. Izvo zvakakosha kuti ticherechedze kuti imwe neimwe katatu mune iyi dataset yakatsanangurwa nevanosvika vatanhatu vanhu.
Iyi dataset inobatsira mapurojekiti anosanganisira kuongororwa kwechiso, sekutaura-kwakavakirwa pikicha kudzoreredza, kupatsanurwa kwemanzwiro, synthesis yekutaura, zvichingodaro. Kuti uwane mukana kune dataset, fomu pfupi rinofanira kuzadzikiswa.
4. Visual Genome
Mubvunzo Wekuona Kupindura dhata munzvimbo ine sarudzo dzakawanda inowanikwa muVisual Genome. Inoumbwa ne101,174 MSCOCO mifananidzo ine 1.7 miriyoni QA pairi, ine avhareji yemibvunzo gumi nenomwe pamufananidzo.
Mukuenzanisa neiyo Visual Mubvunzo Mhinduro yedataset, iyo Visual Genome dhata ine kugovera kwakaringana pamhando dzemibvunzo mitanhatu: Chii, Kupi, Rini, Ndiani, Sei, uye Sei.
Pamusoro pezvo, iyo Visual Genome dhataset inosanganisira 108K mapikicha akaiswa zvakanyanya tag nezvinhu, zvivakwa, uye zvinongedzo.
5. LibriSpeech
Iyo LibriSpeech corpus muunganidzwa weanosvika chiuru chemaawa emabhuku ekuteerera kubva kuLibriVox purojekiti. Mazhinji emabhuku ekuteerera anobva kuProjekti Gutenberg.
Iyo data yekudzidziswa yakakamurwa kuita zvikamu zvitatu zve100hr, 360hr, uye 500hr seti, nepo dev uye data rekuyedza rinenge 5hr mukureba kwekuteerera.
6. The Cityspaces
Imwe yeanonyanya kuzivikanwa mahombe dhatabhesi emavhidhiyo estereo ane maonero emudhorobha anonzi The Cityscapes.
Iine pixel-chaiyo zvirevo zvinosanganisira nzvimbo dzeGPS, tembiricha yekunze, ego-kufamba data, uye ekurudyi stereo maonero, inosanganisira zvakarekodhwa kubva makumi mashanu emaguta eGerman akasiyana.
7. Kinetics Dataset
Imwe yeanonyanya kuzivikanwa vhidhiyo dataset yekuziva zviitiko zvevanhu pamwero mukuru uye nemhando yakanaka ndeye Kinetics dataset. Kune angangoita mazana matanhatu emavhidhiyo ega ega emakirasi mazana matanhatu ezviitiko zvevanhu, anodarika mazana mashanu ezviuru pamwe chete.
Mafirimu akatorwa kubva kuYouTube; imwe neimwe inenge 10 seconds kureba uye ine kirasi imwe chete yebasa yakanyorwa.
8. CelebMask-HQ
CelebAMask-HQ muunganidzwa wemakumi matatu emakumi matatu epamusoro-zviso zviso zvine masiki akanyatsotsanangurwa uye gumi nematatu makirasi anosanganisira zvinhu zvechiso seganda, mhino, maziso, brows, nzeve, muromo, muromo, bvudzi, ngowani, girazi reziso, mhete, mutsipa, mutsipa, zvinhu.
Iyo dataset inogona kushandiswa kuyedza uye kudzidzisa kuzivikanwa kwechiso, kupatsanura kumeso, uye maGAN ekugadzira kumeso uye kugadzirisa algorithms.
9. Penn Treebank
Imwe yeanonyanya kuzivikanwa uye anowanzo shandiswa corpora kuongororwa kwemamodheru ekutevedzana kwemateki ndeye English Penn Treebank (PTB) corpus, kunyanya chikamu chekorasi chinoenderana neWall Street Journal zvinyorwa.
Izwi rega rega rinofanirwa kuve nechikamu chekutaura chakaiswa sechikamu chebasa racho. Chiyero cheunhu uye izwi-mwero mutauro wokuenzanisira zvakare kazhinji inoshandisa corpus.
10. VoxCeleb
VoxCeleb yakakura-hombe yekuzivisa yekutaura dataset inogadzirwa otomatiki kubva open-source media. VoxCeleb ine zvinopfuura miriyoni kutaura kubva kune vanopfuura 6k vatauri.
Sezvo dhatabheti rinosanganisira odhiyo-inotaridzika, inogona kushandiswa kune akasiyana ekuwedzera maapplication, anosanganisira ekuona mataurirwo ekutaura, kupatsanurwa kwekutaura, kuchinjika-modal kutamiswa kubva kuchiso kuenda kune izwi kana zvinopesana, uye kudzidzisa kucherechedzwa kwechiso kubva kuvhidhiyo kuti uwedzere kuzivikanwa kwechiso ikozvino. datasets.
11. SIXray
Iyo dataset yeSIXray inosanganisira 1,059,231 X-ray mifananidzo yakaunganidzwa kubva muzviteshi zvepasi pevhu uye yakatsanangurwa nevanoongorora kuchengetedza kwevanhu kuti vaone marudzi makuru matanhatu ezvinhu zvinorambidzwa: pfuti, mapanga, zvipanera, pliers, chigero, nesando. Uyezve, mabhokisi ekusungirira echinhu chimwe nechimwe chisingabvumirwi akawedzerwa nemaoko kumaseti ekuedzwa kuitira kuti tiongorore kushanda kwenzvimbo yechinhu.
12. Njodzi dzeUS
Zvinhu zvepurojekiti iyi zvakatoburitswa nezita redataset, US Accident. Iyi dataset yenjodzi dzemotokari munyika yose inosanganisira ruzivo kubva muna Kukadzi 2016 kusvika Zvita 2021 uye inovhara matunhu makumi mana nemapfumbamwe muUSA.
Marekodhi etsaona anosvika miriyoni imwe chete nemazana mashanu avepo mumuunganidzwa uyu. Yakaunganidzwa munguva chaiyo nekushandisa akati wandei traffic APIs.
Aya maAPIs anofambisa ruzivo rwetraffic rwakaunganidzwa kubva kwakasiyana siyana, kusanganisira makamera emumigwagwa, masangano ekuchengetedza mutemo, uye US uye nyika madhipatimendi ezvekutakurwa.
13. Ocular Disease Kuzivikanwa
Iyo yakarongwa ophthalmic dhatabhesi yeOcular Disease Intelligent Recognition (ODIR) ine ruzivo rwevarwere zviuru zvishanu, kusanganisira zera ravo, ruvara rwe fundus mumaziso avo ekuruboshwe nekurudyi, uye nyanzvi dzekurapa 'diagnostic keywords.
Iyi dataset muunganidzwa chaiwo wedata revarwere kubva kuzvipatara zvakasiyana-siyana uye nzvimbo dzekurapa muChina iyo Shanggong Medical Technology Co., Ltd. With hutungamiri hwemhando yepamusoro, tsananguro dzakanyorwa nevaverengi vevanhu vane unyanzvi.
14. Chirwere chemoyo
Iri dhata rechirwere cheMwoyo rinobatsira mukuona kuvapo kwechirwere chemoyo mumurwere zvichibva pa76 parameters sezera, murume kana mukadzi, marwadzo emuchipfuva, kuzorora kweropa, zvichingodaro.
Nemakesi e303, dhatabhesi inotsvaga kungosiyanisa kuvepo kwechirwere (kukosha 1,2,3,4) kubva pakusavapo kwayo (kukosha 0).
15. CLEVR
Iyo CLEVR dhata (Compositional Mutauro uye Elementary Visual Reasoning) inotevedzera Visual Mubvunzo Mhinduro. Iyo ine mafoto e-3D-yakashandurwa zvinhu, nepikicha yega yega inoperekedzwa nenhevedzano yemibvunzo yakanyatso kurongeka yakakamurwa mumapoka akati wandei.
Kune ese mapikicha echitima nekusimbisa uye mibvunzo nemibvunzo, dhatabheti rine zviuru makumi manomwe emifananidzo uye zviuru mazana manomwe emibvunzo yekudzidziswa, zviuru gumi nezvishanu mifananidzo uye zviuru zana nemakumi mashanu emibvunzo yekusimbisa, uye zviuru gumi neshanu mifananidzo uye zviuru zana nemakumi mashanu emibvunzo yekuyedza inosanganisira zvinhu, mhinduro, magirafu anoshanda, uye magirafu anoshanda.
16. Universal Dependencies
Iyo Universal Dependencies (UD) chirongwa chine chinangwa chekugadzira muchinjika-mutauro yunifomu morphology uye syntax treebank annotation yemitauro yakawanda. Shanduro 2.7, iyo yakaburitswa muna 2020, ine 183 mabhanga emiti mumitauro zana.
Chirevo ichi chinoumbwa nematagi ePOW epasirese, misoro yekutsamira, uye mazita ekutsamira kwepasirese.
17. KITTI - 360
Imwe yeanonyanya kushandiswa dhataseti yemarobhoti enhare uye kuzvidzivirira kutyaira ndiyo KITTI (Karlsruhe Institute of Technology uye Toyota Technological Institute).
Iyo inoumbwa nemaawa 'akakosha ezviitiko zvetraffic zvakatorwa pachishandiswa huwandu hwema sensor modalities, senge yakakwirira-resolution RGB, grayscale stereo, uye 3D laser scanner kamera. Iyo dataset yakagadziridzwa nekufamba kwenguva nevaongorori vakati wandei vakatsanangura nemaoko zvikamu zvakasiyana-siyana kuti zvienderane nezvavanoda.
18. MOT (Multiple Object Tracking)
MOT (Multiple Object Tracking) idataset yezvakawanda zvekutevera zvinhu izvo zvinosanganisira mukati nekunze zvimiro zvenzvimbo dzeveruzhinji zvinosanganisira vanofamba netsoka sezvinhu zvinofarirwa. Vhidhiyo yechiitiko chega chega yakakamurwa kuita zvidimbu zviviri, chimwe chekudzidziswa uye chimwe chekuyedza.
Iyo dataset inosanganisira kuwanikwa kwezvinhu mumavhidhiyo mafuremu uchishandisa matatu madhigirii: SDP, Faster-RCNN, uye DPM.
19. PASCAL 3D+
Iyo Pascal3D + yakawanda-yekuona dataset inoumbwa nemifananidzo yakaunganidzwa musango, kureva, mifananidzo yezvikamu zvezvinhu zvine musiyano wakanyanya, wakatorwa mumamiriro ezvinhu asingadzoreki, munzvimbo dzakazara vanhu, uye munzvimbo dzakasiyana siyana. Pascal3D+ inosanganisira gumi nembiri dzakaomesesa zvinhu zvikamu zvakatorwa kubva kuPASCAL VOC 12 dataset.
Izvi zvinhu zvine ruzivo rwemamisikirwo akamakwa pazviri (azimuth, kukwira, uye kureba kune kamera). Pascal3D + inosanganisirawo pose-inotsanangurwa mafoto kubva kuImageNet muunganidzwa mune aya gumi nemaviri mapoka.
20. Facial Deformable Models dzeMhuka
Chinangwa cheFacial Deformable Models of Animals (FDMA) chirongwa ndechekupikisa nzira dzazvino mukuzivikanwa kwenzvimbo yechiso chemunhu uye kuronda uye kugadzira maalgorithms matsva anogona kutarisana nekusiyana kwakakura kuri hunhu hwehunhu hwechiso chemhuka.
Algorithms yepurojekiti iyi yakaratidza kugona kuona uye kuteedzera zviso pazviso zvevanhu uchibata nekusiyana kunokonzerwa nekuchinja kwemanzwiro echiso kana zvinzvimbo, kuvharika, uye mwenje.
21. MPII Human Post Dataset
Iyo MPII Human Pose Dataset ine mapikicha anosvika makumi maviri neshanu, makumi maviri neshanu ari masamples ekudzidzira, 25K ayo ari masampula ekusimbisa, uye 15K ayo ari ekuyedza samples.
Zvinzvimbo izvi zvakanyorwa nemaoko anosvika gumi nematanhatu majoini emuviri, uye mafoto anotorwa kubva kumafirimu eYouTube anofukidza mazana mana negumi zviitiko zvevanhu.
22. UCF101
Iyo UCF101 dataset ine zvikamu gumi nezvitatu nemazana matatu nemakumi maviri emavhidhiyo akarongwa muzvikamu zana. Aya 13,320 mapoka akapatsanurwa muzvikamu zvishanu: kufamba kwemuviri, kudyidzana kwevanhu-munhu, kusangana kwevanhu-chinhu, kuridza chiridzwa chemimhanzi, uye mitambo.
Mavhidhiyo acho anobva kuYouTube uye anosanganisira maawa makumi maviri nemanomwe pakureba.
23. Audioset
Audioset idhidhiyo chiitiko dataset inoumbwa nevanhu vanopfuura mamirioni maviri-yakatsanangurwa gumi-yechipiri mavhidhiyo zvikamu. Kutsanangudza iyi data, a hierarchical ontology inosanganisira 2 mhando dzezviitiko inoshandiswa, izvo zvinoreva kuti ruzha rumwe chete runogona kunyorwa zvakasiyana.
24. Stanford Natural Mutauro Inference
Iyo dataset yeSNLI (Stanford Natural Language Inference) ine 570k mitsara miviri yakarongedzerwa nemaoko seyakagadzirirwa, kupokana, kana kwayakarerekera.
Zvivako ndezveFlickr30k tsananguro yemifananidzo, nepo fungidziro dzakagadzirwa nechaunga-chinobva vanyori vakapihwa chivakwa uye vakarairwa kuti vaburitse zvirevo, zvinopokana, uye zvisina kwazvakarerekera.
25. Kupindurwa kwemibvunzo yekuona
Visual Question Answering (VQA) idataset rine mibvunzo yakavhurika maererano nemifananidzo. Kuti upindure mibvunzo iyi, unofanirwa kubata chiono, mutauro, uye pfungwa.
mhedziso
Sezvo kudzidza kwemichina uye hungwaru hwekugadzira (AI) hunowedzera kuwanda mubhizinesi rega rega uye muhupenyu hwedu hwezuva nezuva, ndizvo zvinoitawo huwandu hwezviwanikwa uye ruzivo rwuripo pachinhu ichi.
Akagadzirira-akagadzirwa eruzhinji dhataseti anopa yakanaka yekutanga nzvimbo yekugadzira AI modhi uku ichibvumirawo vane nguva ML programmers kuchengetedza nguva uye kutarisa kune zvimwe zvinhu zvezvirongwa zvavo.
Leave a Reply