Marketplace Data

Flitto Data Marketplace menyediakan dataset pelatihan dan evaluasi AI berlisensi terverifikasi dalam bentuk teks, suara, gambar, dan video, mencakup lebih dari 100 bahasa dan 23 domain khusus. Coba sampel gratis terlebih dahulu, lalu hubungi kami jika tertarik untuk melakukan pembelian.

Jual Data Anda di Flitto

Daftarkan data dengan mudah dan periksa kelayakannya untuk dijual.

  • NEWEvaluation DataMultimodal

    Urdu Nastaliq OCR/VLM Error-Aligned Fine-Tuning Dataset

    This dataset aligns 161 verified text lines from a 1908 lithographed Urdu dictionary with outputs from eight AI models. It includes runtime metadata and word-level error analysis for OCR/VLM fine-tuning and cross-model performance comparison.

    DomainHumanities and Social
    LanguageUrdu
  • NEWEvaluation DataMultimodal

    Licensed Balochi Printed Book Evaluation Benchmark

    This benchmark uses licensed Balochi printed books, including rare lithographs. It provides held-out evaluation data and native-verified ground-truth text to assess OCR/VLM recognition accuracy and analyze errors.

    DomainHumanities and Social
    Language-
  • NEWEvaluation DataMultimodal

    Arabic Historical Manuscript & Early Print Evaluation Benchmark

    This Arabic OCR/VLM benchmark provides held-out evaluation data, expert-verified transcriptions, and model-specific error analysis for rare manuscripts and early prints with handwriting and marginal notes. Existing models showed up to 96% word-recognition discrepancies.

    DomainHumanities and Social
    LanguageArabic
  • NEWEvaluation DataMultimodal

    Archival Persian Nastaliq Lithograph Evaluation Benchmark

    This Persian Nastaliq OCR/VLM benchmark provides held-out evaluation data, expert-verified transcriptions, and model-specific error analysis. It assesses recognition accuracy on early 20th-century lithographs with reported word-recognition discrepancies of up to 76%.

    DomainHumanities and Social
    LanguagePersian
  • NEWFine-tuning DataMultimodal

    Licensed Balochi Printed Book Corpus

    This licensed Balochi print corpus includes rare Arabic-script lithographs, original page scans, OCR text, and native-expert verified transcriptions. It supports OCR and VLM training and fine-tuning for Balochi text recognition.

    DomainHumanities and Social
    Language-
  • NEWFine-tuning DataMultimodal

    Archival Arabic Manuscript & Early Print Ground-Truth Corpus

    This dataset provides expert-verified transcriptions of rare Arabic manuscripts and early prints with handwriting and marginal notes. OCR/VLM models showed up to 96% word-recognition discrepancies. It supports model training and fine-tuning for complex Arabic text recognition.

    DomainHumanities and Social
    LanguageArabic
  • NEWFine-tuning DataMultimodal

    Archival Persian Nastaliq Lithograph Ground-Truth Corpus

    This dataset provides expert-verified Persian Nastaliq transcriptions from rare early 20th-century lithographs. OCR/VLM models showed up to 76% word-recognition discrepancies. It supports training and fine-tuning to improve Persian text recognition.

    DomainHumanities and Social
    LanguagePersian
  • NEWEvaluation DataMultimodal

    Licensed Pashto Print OCR Evaluation Benchmark

    An OCR evaluation dataset built from licensed Pashto printed books. It provides an evaluation set separate from the training corpus, allowing OCR recognition performance on printed Pashto documents to be assessed against native-verified reference transcriptions.

    DomainHumanities and Social
    LanguagePashto
  • NEWPre-training DataText

    Major Website Custom Crawling & Dataset Construction

    This custom dataset is built by collecting and processing web content from selected websites based on project requirements. It supports AI model development and data analysis.

    DomainIT and Tech
    Language-
  • NEWFine-tuning DataMultimodal

    Verified Professional Video Interview & Identity Ground-Truth Dataset

    This multimodal ground-truth dataset consists of recorded professional interviews with identity-verified candidates who have provided explicit data-licensing consent. Each record pairs interview video and unscripted transcripts with anonymized CVs, liveness checks, and behavioral integrity flags acr

    DomainManagement, Economic and Finance|Science and Engineering
    Language-