Contents x
    Specifications
    • 27 Oct 2025
    • Print
    • Dark
      Light
    • PDF
    Contents

    Specifications

    • Updated On 27 Oct 2025
    • Print
    • Dark
      Light
    • PDF
    Article summary

    This page provides insights into the prerequisites and criteria necessary for utilizing the different functionalities of the Dataloop platform.

    Recommended PC/Workstation

    Ensure smooth performance by using a workstation that meets the following recommended specifications:

    Type

    Recommended

    RAM

    8GB

    CPU

    Intel(R) Pentium(R)3805@ 1.90GHz x 2 Processor 64 bit

    Screen resolution

    1920 x 1080

    Internet connection

    Speed test: Download:18.61 Mbps / Upload: 15.68 Mbps / Ping: 31 ms

    Browser Requirements

    Access the Dataloop console using one of the supported browsers. We recommend using Google Chrome for the best experience:

    Supported Browser

    Minimum Version

    Chrome

    47

    Safari

    9

    Firefox

    45

    Opera

    36

    Edge

    12

    Support

    If you encounter issues using the web interface with any of the listed browsers, please contact Dataloop Support.

    Recipe and Ontology

    Configure ontologies and labels efficiently within these recommended limits:

    Type

    Recommended

    Labels

    400 labels per ontology

    Attributes

    100 attributes per ontology

    Data Management

    Ensure dataset scalability and performance with the following limitations:

    Type

    Limit

    Annotation label

    1024 chars

    Annotation attributes

    Max 100 per annotation

    Annotations in joined queries

    Maximum 7,000,000 items

    Items in the Dataset

    Maximum 25,000,000 items

    Item JSON file size

    Maximum 100 KB

    Item filename (full path including dirs)

    1024 chars

    Maximum items allowed when cloning a dataset via UI

    50,000 items

    Maximum download file size from the platform

    2 GB

    Schema enforcement in Dataset

    When schema enforcement is enabled on a dataset, Dataloop requires that each metadata key (path) maintains a consistent data type across all items.

    For example, if metadata.user.device_id is initially set as a string (e.g., "cam_01"), you must not assign a different type such as an integer (e.g., 123) to the same key elsewhere in the dataset.

    Search size per each search key/value

    1024 Bytes

    Unique custom metadata keys on items in dataset

    • Searchable keys: Maximum is 100 searchable keys per dataset under metadata.user. To add more keys and make them unsearchable, refer to the Unsearchable Metadata article.

    • Unsearchable keys: Maximum of 200 searchable keys per dataset

    Queries limitations

    Understand system constraints when constructing queries:

    Query

    Limit

    Maximum total filter terms in a query

    20 terms

    Maximum metadata filter terms in a query

    10 terms

    Max sort terms in a query

    3 terms

    Sorting over custom metadata

    1 term

    Filter by annotation Metadata

    Not supported

    Notice

    • Dataloop permits the storage and querying of over 25,000,000 items and unlimited annotations in a dataset, However, join queries between items and annotations are limited to searching over max 7M annotations.

    • A Term refers to an individual element or condition used in the query to filter, sort, or search for data. Each term typically represents a distinct criterion or parameter that helps refine the results of a query.

    Supported File Formats

    Dataloop supports a wide range of file formats across domains to ensure flexible and scalable annotation workflows.

    Type

    Recommended

    Image

    JPG, JPEG, PNG, and TIFF

    Video

    WEBM, MP4, and MOV

    Audio

    WAV, MP3, OGG, FLAC, M4A, and AAC

    Lidar

    PCD (ASCII format with a size limit of 70 MB)

    NLP / NER

    TXT, JSON, EML, and PDF

    GIS

    COG, GeoTIFF, OSM, and XYZ

    Notice

    • Audio Studio: Convert FLAC to MP3 to reduce file size without significant quality loss.

    • Video Studio:WEBM is the default format for video annotation. For more information, see the WEBM article.

    • MP4 Limitations: SAR/DAR encoding values are not supported in MP4 files.

    Image Item

    Define optimal conditions for image annotation:

    Type

    Recommended

    Size

    15 MB

    Formats

    Refer to the Supported File Formats section.

    Resolution

    24.0 MP

    Total annotations

    1,000

    Important

    Valid EXIF values are 1 to 8. Values outside this range (for example, 0) are ignored. Read more about Exif orientation values.

    Audio Item

    Best practices for audio file handling and annotation:

    Type

    Recommended

    Annotations

    Up to 300

    Recommended format

    Refer to the Supported File Formats section.

    Recommended audio size

    Up to 100 MB

    Important

    File size depends on bitrate, compression, and encoding, not just duration. A 17-minute file can be larger than a 1-hour file if encoded with higher fidelity.

    Video Item

    Guidelines for working with video files in annotation projects:

    Type

    Recommended

    Recommended format

    Refer to the Supported File Formats section.

    Recommended max video size

    150 MB

    Total annotations

    Up to 400

    Task

    Control and optimize task assignment across annotation workflows:

    Type

    Recommended

    Maximum unworked items per task (without Status)

    80,000

    Maximum assignees per task

    1,000

    Maximum tasks per project

    5,000

    Assignment Limits

    Define assignment structure within large-scale tasks:

    Type

    Recommended

    Maximum items per single assignment

    10,000

    Annotation Studio

    Annotation Studio supports advanced visualization and configuration:

    Type

    Definition

    Recommended

    Multi-Modality layers

    Combine multiple data types (e.g., image + text + metadata) in a single view.

    8

    Similarity items

    Used for similarity evaluation workflows (e.g., GenAI outputs).

    8

    Multi-view items

    Visualize multiple perspectives or sequential data layers for richer context.

    6

    What's Next