I will be presenting two of our projects in the ECCV 2026 main track this week, and one of them is a spotlight!
The spotlight is sparse to dense prediction, for cloud properties.
1
3
187
The second paper is a foundation model for the Moon. Seven modalities from five NASA and ISRO missions aligned onto one grid, so a masked autoencoder learns a single representation across all of them. Beats ImageNet-pretrained and from-scratch on all six tasks.
1
1
34
Ice Cloud Geometry Retrieval with Calibrated Uncertainty from Passive Satellite Imagery
ayushprasad.com/projects/spa…
Moonstone: A Multimodal Foundation Model and Benchmark for Lunar Remote Sensing
ayushprasad.com/projects/moo…
25
Ayush Prasad retweeted
Recently published:
Modeling snow on sea ice using physics-guided machine learning
👉 bit.ly/3PAFAwX
By @ayushprd, @IoMerk & @AleksiNummelin
#emulator #physicsguidedml #MachineLearning #SnowModel #SnowOnSeaIce #ClimateChange #ClimateModelling
Ayush Prasad retweeted
What drives innovation in machine learning?
In a new paper, we argue that application-driven work is systemically under-valued in the machine learning community, but that it's essential for both innovation and impact.
arxiv.org/abs/2403.17381
1/3
ALT A figure with two columns. Under "Methods-driven ML" are the attributes "stereotyped benchmarks", "narrow set of evaluation metrics", "massive datasets", and "problem-agnostic methods". Under "Application-driven-ML" are the attributes "real-world tasks", "application-specific evaluation metrics", "auxiliary domain knowledge", and "use-inspired methods".
Ayush Prasad retweeted
BREAKING: Airbus announces deal with Indian LCC Indigo for 500 A320neo family aircraft, the largest single deal in aviation history.
Excited to present our work on explainable machine learning for marine heatwaves forecasting at #EGU23 🌊🌡️ – at Vienna, Austria
Ayush Prasad retweeted
Known for shedding light on the AI model, XAI has the potential to unveil new knowledge for applications in climate science – but which of the many XAI methods can we trust and should we choose?
See our preprint
arxiv.org/pdf/2303.00652.pdf
@TUBerlin_UMI with @Marlene_Climate
Ayush Prasad retweeted
A biotech company has announced an audacious effort to ‘de-extinct’ the dodo. Could the plan actually work? go.nature.com/3Iudwsk
Ayush Prasad retweeted
It's been almost ~5yrs since I returned to India after 15yrs in the US. I've written about it before but here is my story 5yrs later. It's been a ride & frankly one of my life's best decisions! If you are considering a move back to India, hope this helps:
medium.com/@MohapatraHemant/…
Ayush Prasad retweeted
Continuing the conversation on forest carbon credits quality, @PachamaInc just published a Research Brief, outlining our tech-driven approach to computing robust algorithmic baselines with validated uncertainty for avoided deforestation crediting.
pachama.com/blog/pachama-res…
Ayush Prasad retweeted
Super excited to present (albeit virtually) our @manpa_97 work on Mapping Biogeophysical Parameters Using Satellite Data and Geospatial Technologies at #AGU22. Drop by to hear me talk about it on Wed 14 at 20:55!
@theAGU @ClimateResearch @HariniNagendra
What an inspiring and exciting day it was at Slush 2022! Hearing about startups solving a wide range of problems from hyperspectral remote sensing to climate-friendly neo banks, kudos to the @SlushHQ team! #slush2022 – at Helsinki, Finland
1
3
Ayush Prasad retweeted
In 2018, three colleagues and I published a front-cover paper in @PNASNews
It has > 170 citations as of 2022, and it was a success that might not have happened:
• We had $0 budget
• We had no PI (Principal Investigator)
• We had no credentials
[Here's how we did it] 🧵
Ayush Prasad retweeted
At today's @ImageUCPH workshop, section head @KErleben asked me to speak about "how to get cool research ideas." (1/25)
Nice comprehensive overview of various explainable AI methods link.springer.com/chapter/10…
1
Ayush Prasad retweeted
flew first time with @Finnair E190! absolute loved the flight 🛫 from crew to food to interiors all 💯
