james-burton/OrientalMuseum_min3-3Dwhite-name

--- dataset_info: features: - name: obj_num dtype: string - name: file dtype: string - name: image dtype: image - name: root dtype: string - name: description dtype: string - name: label dtype: class_label: names: '0': Aegis '1': Ajaeng Holder '2': Album Painting '3': Amulet Mould '4': Animal Figurine '5': Animal Mummy '6': Animal bone '7': Arm Guard '8': Axe Head '9': Axle-caps '10': Ball '11': Ballista Bolt '12': Band '13': Basin '14': Baton '15': Bead Net '16': Belt Hook '17': Betel Nut Cutter '18': Blouse '19': Blu-ray disc '20': Bolt '21': Book Cover '22': Box '23': Brush Pot '24': Brush Rest '25': Brush Tray '26': Bulb Bowl '27': Bullet Mould '28': Burnisher '29': Cabinet '30': Cannon '31': Cap '32': Carved stone '33': Case '34': Cash Box '35': Chest '36': Cigar Holder '37': Clapper '38': Clay pipe (smoking) '39': Comb '40': Compass '41': Cosmetic and Medical Equipment and Implements '42': Counterpoise '43': Cricket pot '44': Cross-bow Lock '45': Cup And Saucer '46': Cup, Saucer '47': Cushion Cover '48': DVDs '49': Dagger '50': Dice Box '51': Dice Shaker '52': Disc '53': Domestic Equipment and Utensils '54': Double Dagger '55': Dummy '56': Ear Protector '57': Ear Stud '58': Earring '59': Elephant Goad '60': Erotic Figurine '61': Eye Protector '62': Fan Case '63': Feet Protector '64': Ferrous object '65': Figurine Mould '66': File '67': Finger Ring '68': Fitting '69': Flannel '70': Flute '71': Funerary Cone '72': Funerary goods '73': Funerary money '74': Furosode '75': Greek crosses '76': Hand Jade '77': Hand Protector '78': Handwarmer '79': Hanging '80': Headband '81': Heart Scarab '82': Human Figurine '83': Incense Holder '84': Inkstick '85': Jue (jade) '86': Kite '87': Knee Protector '88': Kohl Pot '89': Kundika '90': Leaflet '91': Leg '92': Leg Protector '93': Letter '94': Lock '95': Mah Jong Rack '96': Majiang set '97': Manuscript Page '98': Massager '99': Mat '100': Mica Painting '101': Miniature Painting '102': Miniature Portrait '103': Mortar '104': Mould '105': Mouth Jade '106': Mouth Protector '107': Mouth-piece '108': Mummy Label '109': Nail Protector '110': Neck Guard '111': Nose Protector '112': Opium Pipe '113': Opium Weight '114': Oracle Bone '115': Ostraka '116': Paddle '117': Palette '118': Panel '119': Part '120': Pelmet '121': Pencase '122': Pendant '123': Perfumer '124': Phallus Protector '125': Phylactery '126': Pigstick '127': Pipe '128': Pipe Case '129': Pipe Holder '130': Pith Painting '131': Plaque '132': Plate '133': Poh Kam '134': Pounder '135': Prayer Wheel '136': Quoit '137': Rank Square '138': Rubber '139': Sake Cup '140': Scabbard Chape '141': Scabbard Slide '142': Scarab Seal '143': Scarf '144': Score Board '145': Screen '146': Seal '147': Seal Paste Pot '148': Shaft Terminal '149': Shield '150': Shroud Weight '151': Sleeve Band '152': Sleeve Weight '153': Slide '154': Soles '155': Spillikins '156': Staff Head '157': Stamp '158': Stand '159': Stand of Incense Burner '160': Stem Bowl '161': Stem Cup '162': Story Cloth '163': Strainer '164': Sword Guard '165': Sword Knob '166': T-shirts '167': Table '168': Table Runner '169': Thangka '170': Throwing Stick '171': Tomb Figure '172': Tomb Model '173': Tongue Protector '174': Washer '175': Water Dropper '176': Water Pot '177': Wine Pot '178': Womb Protector '179': Woodblock Print '180': Writing Desk '181': accessories '182': adzes '183': alabastra '184': albums '185': altar components '186': altars '187': amphorae '188': amulets '189': anchors '190': animation cels '191': animation drawings '192': anklets '193': armbands '194': armor '195': armrests '196': arrowheads '197': arrows '198': autograph albums '199': axes '200': 'axes: woodworking tools' '201': back scratchers '202': badges '203': bags '204': balances '205': bandages '206': bangles '207': banners '208': baskets '209': beads '210': beakers '211': bedspreads '212': bells '213': belts '214': bezels '215': bi '216': blades '217': blowguns '218': board games '219': boats '220': boilers '221': bone '222': booklets '223': books '224': bottles '225': bowls '226': boxes '227': bracelets '228': bread '229': brick '230': brooches '231': brush washers '232': brushes '233': buckets '234': buckles '235': business cards '236': buttons '237': caddies '238': calendars '239': calligraphy '240': candelabras '241': candleholders '242': candlesticks '243': canopic jars '244': card cases '245': card tables '246': cards '247': carvings '248': cases '249': cash '250': celestial globes '251': censers '252': chains '253': chairs '254': charms '255': charts '256': chess sets '257': chessmen '258': chisels '259': chokers '260': chopsticks '261': cigarette cases '262': cigarette holders '263': cippi '264': clamps '265': clappers '266': claypipe '267': cloth '268': clothing '269': coats '270': coffins '271': coins '272': collar '273': combs '274': compact discs '275': containers '276': coverings '277': covers '278': crucifixes '279': cuffs '280': cups '281': cushions '282': cutlery '283': cylinder seals '284': deels '285': deity figurine '286': diagrams '287': dice '288': dishes '289': document containers '290': documents '291': dolls '292': doors '293': drawings '294': dresses '295': dressing gowns '296': drums '297': dung-chen '298': earrings '299': embroidery '300': ensembles '301': envelopes '302': 'equipment for personal use: grooming, hygiene and health care' '303': ewers '304': fans '305': fasteners '306': 'feet: furniture components' '307': female figurine '308': ferrules '309': fiddles '310': figures '311': figurines '312': finials '313': fishhooks '314': flagons '315': flags '316': flasks '317': flint '318': fragments '319': funnels '320': furniture components '321': gameboards '322': games '323': gaming counters '324': ge '325': glassware '326': gloves '327': goblets '328': gongs '329': gowns '330': greeting cards '331': hair ornaments '332': hairpins '333': hammerstones '334': handkerchiefs '335': handles '336': handscrolls '337': hanging scrolls '338': harnesses '339': hatpins '340': hats '341': headdresses '342': headrests '343': heads '344': headscarves '345': helmets '346': hobs '347': hoods '348': hooks '349': houses '350': identity cards '351': illuminated manuscripts '352': incense burners '353': incense sticks '354': ink bottles '355': inkstands '356': inkstones '357': inkwells '358': inlays '359': iron '360': jackets '361': jar seal '362': jars '363': jewelry '364': jue '365': juglets '366': jugs '367': kayagum '368': keys '369': kimonos '370': knives '371': kŏmun'gos '372': ladles '373': lamps '374': lanterns '375': lanyards '376': leatherwork '377': lids '378': lockets '379': loom weights '380': maces '381': manuscripts '382': maps '383': maquettes '384': masks '385': medals '386': miniatures '387': mirrors '388': miscellaneous '389': models '390': money '391': mortarboards '392': mounts '393': mugs '394': mummies '395': musical instruments '396': nails '397': necklaces '398': needles '399': netsukes '400': nozzles '401': obelisks '402': obis '403': oboes '404': oil lamps '405': ornaments '406': overdresses '407': pages '408': paintings '409': paper money '410': paperweights '411': papyrus '412': passports '413': pectorals '414': pendants '415': pennants '416': pestles '417': petticoats '418': photograph albums '419': photographs '420': pictures '421': pins '422': pipes '423': pitchers '424': plaques '425': plaster '426': playing card boxes '427': playing cards '428': plinths '429': plumb bobs '430': plumbing fixtures '431': plume holders '432': poker '433': pommels '434': postage stamps '435': postcards '436': posters '437': pots '438': pottery '439': prayer beads '440': prayers '441': printing blocks '442': printing plates '443': prints '444': punch bowls '445': puppets '446': purses '447': puzzles '448': pyxides '449': quilts '450': rag-dung '451': razors '452': reliefs '453': rifles '454': rings '455': robes '456': roofing tile '457': rosaries '458': rose bowls '459': rubbings '460': rugs '461': rulers '462': sandals '463': saris '464': sarongs '465': sashes '466': sauceboats '467': saucers '468': saws '469': scabbards '470': scaraboids '471': scarabs '472': scarves '473': scepters '474': scissors '475': scrolls '476': sculpture '477': seed '478': seppa '479': shadow puppets '480': shawls '481': shears '482': shell '483': shelves '484': sherds '485': shields '486': shoes '487': shrines '488': sistra '489': situlae '490': sketches '491': skewers '492': skirts '493': snuff bottles '494': socks '495': spatulas '496': spearheads '497': spears '498': spittoons '499': spoons '500': stampers '501': staples '502': statues '503': statuettes '504': steelyards '505': stelae '506': sticks '507': stirrup jars '508': stools '509': stoppers '510': straps '511': studs '512': styluses '513': sugar bowls '514': sugar tongs '515': swagger sticks '516': swords '517': tablecloths '518': tablets '519': tacks '520': talismans '521': tallies '522': tangrams '523': tankards '524': tea bowls '525': tea caddies '526': tea kettles '527': teacups '528': teapots '529': telephones '530': ties '531': tiles '532': toggles '533': toilet caskets '534': tools '535': toys '536': trays '537': trimming '538': trophies '539': trousers '540': trumpets '541': tubes '542': tureens '543': tweezers '544': typewriters '545': underdresses '546': underwear '547': unidentified '548': urinals '549': ushabti '550': utensils '551': vases '552': veils '553': vessels '554': votive offerings '555': waistcoats '556': wall tile '557': watches '558': weighing devices '559': weight '560': weights '561': whetstones '562': whistles '563': whorls '564': wire '565': wood blocks '566': writing boards '567': xylophones - name: other_name dtype: string - name: material dtype: string - name: production.period dtype: string - name: production.place dtype: string splits: - name: validation num_bytes: 689672761.257 num_examples: 5489 - name: test num_bytes: 658367756.568 num_examples: 5489 - name: train num_bytes: 4933427063.75 num_examples: 116625 download_size: 6297757734 dataset_size: 6281467581.575 configs: - config_name: default data_files: - split: validation path: data/validation-* - split: test path: data/test-* - split: train path: data/train-* ---

原始信息汇总

数据集概述

数据集特征

• obj_num: 字符串类型
• file: 字符串类型
• image: 图像类型
• root: 字符串类型
• description: 字符串类型
• label: 分类标签，包含多个类别，例如：
  • 0: Aegis
  • 1: Ajaeng Holder
  • 2: Album Painting
  • ...
  • 566: writing boards
  • 567: xylophones
• other_name: 字符串类型
• material: 字符串类型
• production.period: 字符串类型
• production.place: 字符串类型

数据集分割

• validation: 包含5489个样本，大小为689672761.257字节
• test: 包含5489个样本，大小为658367756.568字节
• train: 包含116625个样本，大小为4933427063.75字节

数据集大小

• 下载大小: 6297757734字节
• 数据集大小: 6281467581.575字节

配置

• config_name: default
  • data_files:
    • validation: data/validation-*
    • test: data/test-*
    • train: data/train-*

搜集汇总

数据集介绍

构建方式

该数据集源自东方博物馆的馆藏资源，经过精心的数字化采集与整理而成。构建过程中，首先对海量文物进行三维扫描与拍摄，获取其立体模型与平面影像数据。随后，依据文物类型、材质、年代、出土地点等多元属性进行系统化标注，形成结构化的标签体系。数据集涵盖了从史前陶器到近现代工艺品的广泛类别，共计567个细分类别，每个样本均包含编号、文件路径、图像、描述文本、标签、别名、材质、生产时期及地点等多维信息。最终，数据被划分为训练集、验证集和测试集三个子集，以支持深度学习模型的训练与评估。

使用方法

使用该数据集时，研究者可通过HuggingFace Datasets库直接加载，指定配置名'default'即可获取train、validation、test三个划分。数据以字典形式返回，包含'obj_num'、'file'、'image'、'root'、'description'、'label'、'other_name'、'material'、'production.period'、'production.place'等字段，其中'image'为PIL图像对象，'label'为整数编码的类别索引。建议利用'description'字段进行多模态对齐学习，或结合'material'与'production.place'开展文化背景下的分类任务。对于三维重建任务，可借助'file'路径读取对应的3D模型文件。数据集的批量处理与缓存机制确保了高效迭代，适用于PyTorch、TensorFlow等主流框架。

背景与挑战

背景概述

在文化遗产数字化浪潮中，博物馆藏品的精确分类与识别成为计算机视觉与数字人文交叉领域的研究热点。由James Burton主导构建的OrientalMuseum_min3-3Dwhite-name数据集，于近年发布，旨在为东方博物馆藏品提供标准化、细粒度的三维影像与多模态标注资源。该数据集涵盖了从史前器物到近代工艺品的广泛类别，总计567个细分类别，包括Aegis、Oracle Bone、Thangka等极具文化特异性的文物类型，并附有材质、产地、年代等结构化元数据。其核心研究问题聚焦于如何利用深度学习模型在高度类间相似性与类内多样性的条件下实现精准文物识别，为文化遗产的自动化编目、虚拟展览与跨库检索提供了关键基准，推动了领域内从粗粒度分类向细粒度语义理解的范式演进。

当前挑战

该数据集面临的核心挑战源于文物识别领域的固有复杂性。首先，类别间视觉歧义性极为突出，例如同属容器类的Basin、Bowl与Jar在形态上高度相似，而不同文化背景下的同类器物（如Chinese jue与Korean jue）又存在微小差异，导致模型易产生混淆。其次，文物三维影像的构建过程存在显著困难，包括多视角光照条件不一致、表面纹理退化（如青铜器锈蚀与织物褪色）以及透明/半透明材质（如玉器与玻璃器）的反射干扰，这些因素加剧了特征提取的难度。此外，数据标注面临历史知识断层挑战，部分文物（如unidentified类别）因缺乏可靠文献佐证而难以准确归类，且跨文化术语体系（如西方‘amphorae’与东方‘jar’的边界模糊）进一步增加了标签噪声，制约了监督学习方法的泛化性能。

常用场景

经典使用场景

OrientalMuseum_min3-3Dwhite-name 数据集以东方博物馆馆藏文物为依托，涵盖从古代兵器、祭祀用具到日常器皿、服饰配饰等数百种精细分类的三维白模图像与元数据，为计算机视觉领域的细粒度文物识别与分类任务提供了极具挑战性的基准。研究者可借助该数据集训练深度神经网络模型，在高度相似的文物形态间进行精准区分，例如区分不同文化背景下的‘碗’与‘钵’，或辨识功能相近但形制迥异的‘剑’与‘匕首’，从而推动面向文化遗产的细粒度图像理解技术发展。

解决学术问题

该数据集有效解决了文化遗产数字化研究中长期存在的两大瓶颈：一是缺乏大规模、多类别、标准化标注的三维文物图像资源，导致跨文化比较与自动分类模型训练困难；二是文物类别间存在类间相似度高、类内差异大的特性，传统分类方法难以胜任。通过提供包含116,625个训练样本、涵盖567个精细类别的白模图像与结构化描述，它为学术社区构建了评估细粒度视觉识别算法鲁棒性与泛化能力的理想平台，进而促进了迁移学习、少样本学习等方法在文化遗产领域的理论突破。

实际应用

在实际应用中，该数据集可赋能博物馆智慧化管理系统，例如自动对入库文物进行初步分类与标签生成，辅助策展人高效组织专题展览；同时，它也可集成到移动导览应用中，允许参观者通过拍摄文物照片即时获取其历史背景、材质与制作工艺等信息。此外，基于该数据集训练的模型还可用于网络文物交易平台的图像审核，自动识别与过滤非法或可疑文物，助力文化遗产保护与监管。

数据集最近研究

最新研究方向

该数据集聚焦于东方博物馆藏品的多模态细粒度分类与数字化保护研究，涵盖超过560种文物类别，如甲骨、青铜器、陶瓷、织物及雕塑等，为文化遗产的智能化识别与知识图谱构建提供了坚实基础。当前前沿方向包括基于深度学习的文物自动鉴定与年代推断、跨文化对比分析以及3D模型重建与虚拟展示技术。结合数字人文热潮，该数据集推动了文化遗产的全球可访问性与教育应用，尤其在非西方艺术品的计算分析中填补了数据空白，对促进跨学科研究与文化多样性保护具有深远意义。