SPIE Prism Award 2023 Winner
Privacy is important to us, so you have the option of disabling certain types of storage that may not be necessary for the basic functioning of the website. Blocking categories may impact your experience on the website. More information
Projected to be produced by a digital pathology department of one hospital
User-level data published from Sentinel satellites, with approximately 230TB of data downloaded daily
time spent by data scientists preparing data rather than creating insights
Generated by a test autonomous vehicle
Errors bounded (single pixel)
Errors independent
No artefacts
Embedded noise model
Compression ratio, avg.
Compression speed, avg.
Lossy
compression
JPEG
7:1
100MB/s
per core
Limited
error
BD3, LERC
5:1
50MB/s
per core
Visually
lossless
JPEG-2000
7:1
10MB/s
per core
Lossless
compression
LZW, FLIF
1.5:1
10MB/s
per core
Metrologically
accurate
Dotphoton
7:1
200MB/s
per core
Cost of cloud storage
& network, $/year
CO2 emissions,
tonnes/year
Time required
to transfer on 100Gbps
8:1
typical
compression
ratio
business
Retain & reuse more data for legal & publication requirements and increase research efficiency
business
More resources for research
Free up time by transmitting petabytes of image data 8x faster on cloud or local storage
business
Less CO2 emissions
Reduce carbon footprint 8x by lowering storage needs and energy consumption with smaller data
business
Equally good for human and machine vision
Ensure the metrological quality of images based on our embedded noise model & sensor data
ML
Laboratory -> Clinical
Establish requirements & tolerances moving from research to production instruments
business
Increase model generalisation
by propagating errors with Monte-Carlo simulations on synthetic data emulating different acquisition modalities
ML
Faster cloud processing
Cloud-based image handling at low latency for more reliable and scalable data-centric AI for connected and autonomous vehicles
business
Identify out-of-distribution data
Discover ML algorithm tolerances to instrument & environmental changes, e.g. resolution, temperature, lighting, focus, contrast
ML
84%
less energy used with networking & storage optimization
business
Normalize data to one virtual sensor
Higher reliability & lower latency from a more compact model by normalising data
ML
6:1
typical compression ratio
business
Higher resolution & better AI
Add quality validation without increasing hardware costs related to RAM, storage and in-car communications
ML
Higher throughput
over light and affordable cable connections
ML
Reduce prototyping
Discover how satellite optical & sensor parameters affect AI/ML performance models to enhance their robustness
ML
Increase satellite sensor resolution & frame rate
Work with higher quality images applying ML super-resolution and denoising
ML
4x
more images stored & processed on the onboard RAM and flash
business
4:1
typical compression ratio
business
Download x4 more images on the same downlink
business
Lower costs & CO2 emissions
Distribute and store earth observation image data on ground faster, more affordably and environmentally-efficient
business
Generate synthetic image data from auto-labelled aerial & drone images to achieve ML results with better-than-human performance
ML
5:1
more images on cloud without plan upgrade
Keep memories, not the CO2
Access/store image data more affordably & environmentally-efficient
55+ cameras supported
No processing limitations
No dynamic range change
No color loss
No bit count change
Save hundreds on new HDD/SSD and NAS upgrades
Jetraw Core
Fast & power-efficient FPGA compression for camera manufacturers
Used by imagery experts
Roberto Camarero,
Onboard Payload Data Processing Engineer European Space Agency
Physical data models in machine learning imaging pipelines
Machine Learning and the Physical Sciences workshop, NeurIPS 2022
Data models for dataset drift controls in machine learning with images
Preprint available (2022)
Data-centric AI workflow based on compressed raw images
8th International Workshop on On-Board Payload, Athens, 26 September 2022
ML4H auditing: from paper to practice
Proceedings of the Machine Learning for Health, PMLR 136:280-317, 2020
