91% of teams take > 3 weeks to create their first dataset.
And months to get to Beta.
Here’s why: The process of creating datasets gets blocked by default.
Teams create many sets. This is because the needed abstractions (such as various Templates) only become known through an iterative process.
With Diffgram, your Data Science team moves into a higher level control position of the Datasets.
The system manages the data, including the Sync process between files.
This makes the process non-blocking, and improves your time spent on the process by up to 3x.
Use Cases
* Create, Update, And Maintain Datasets
* Create Processes for working with Deep Learning systems
* Launch faster, Control costs, Reduce engineering burden
The Sync Engine – Patent Pending
* Event driven data sync. Both at external boundaries and internal sets.
* Create non-linear data flows on demand.
* Dynamically manage complex scenarios, such as multiple sets relating to each other, or conditional relationships.
Runs on Your Hardware (or Hybrid Cloud)
* Easy Setup with Docker
* User Controllable Updates. Code Audits
* Compatible with your cloud system and ML framework.
Controller Agnostic (User Interfaces & AI Assist)
* Use Diffgram, Datasuar, Scale, Super Annotate, or Labelbox,
* Use with your existing or new in house Controller
* Mix and match as needed – complimentary to your system
Best in Class Image & Video Controller
* Object Detection, Segmentation, Classification
* Special support for long videos, including 10 hour+ streams
* Video works with all tasks types, all attribute types, and all spatial types (competitors often restrict what types work with video)
* Track objects, multiple Sub Sequences per video
* High FPS video, with 18,000/frames/clip ( >10x nearest competitor)
* High Resolution Video, up to 4k
Collaborate like never before!
* Engage multiple stakeholders. Including multiple levels of expertise.
* Deep linking to easily share exactly where you are.
* Automatic task distribution and Quality Control functions
* Achieve the highest quality training data.
* Manage people, quality, and process all in one.
*Best in Class Pre-Label Support Use your models to improve performance.
Every feature you need.
Over 100 major features, here’s a preview:
* Core: Multiple Users, Role-Based Access Control, History of Activity, Multiple Permission Groups
*Data: Event Driven Sync, Data Orchestration, Datasets, Relational *Datasets, Create & Update Sets, Data Version Control and Training Data Management.
*Tasks: Control Work, Work with Multiple Controllers
Trusted by Thousands.
*Most referenced Training Data Software in literature (2020)
100% YTD (2020) enterprise retention
*Over 1,000 projects, and 1.5 million instances created on the shared platform
*Users from top companies include EY, Google, Walmart Labs, and Amazon.
*Visual Controller: Box (Object Detection), Polygons, Lines, Keypoints, Classification, Segmentation, Attributes (Nested, Multiple Select, Free Text)
*Quality: Review Tasks, Performance Tracking, Consensus
*Integrations & Extensions: Python SDK, API, Deep existing integrations with AWS, GCP and Azure.
Over 100 pages of Quality Documentation
*Explains concepts – not just specifics
*Many examples and references integrated directly into the application.
Affordable – Starting at less then 1/3 an Engineer.
>12x lower cost then 3 person team
*Month to Month Contracts
*Simple upfront all in pricing
*Grow your Datasets with Diffgram
*Create novel Datasets in Diffgram
Iterate on existing Datasets to grow them.
Create many interlinking Datasets
Enterprise Grade
* Technical support where you need it, when you need it.
* You own and control your data.
* Dedicated point of contact with direct video support.
* Integration and onboarding support.
* Invoice billing.
