Accurate object localization at reduced annotation/computational cost?
#TriLite achieves this by pairing a pre-trained backbone, with our #TriHead component which separates ambiguous, foreground and background features.
The main highlights of #TriLite include:
โข State-of-the-art performance
โข < 1M trainable parameters
โข single-stage training
We will be glad to get in touch at #CVPR2026 to discuss further.
We thank #DEFRA for supporting this research
What a week! ๐
Our group is celebrating a "triple win" with three papers accepted at #ACL2026, #CVPR2026 Workshop, and #FG2026 !
๐๐๐
๐๐ณ๐ฐ๐๐ฐ๐: ๐๐ณ๐ฐ๐ฎ๐ฐ๐ต๐ช๐ฏ๐จ ๐๐ณ๐ฐ๐ด๐ฐ๐ค๐ช๐ข๐ญ ๐๐ฆ๐ฉ๐ข๐ท๐ช๐ฐ๐ถ๐ณ ๐ท๐ช๐ข ๐๐ฉ๐ฆ๐ฐ๐ณ๐บ ๐ฐ๐ง ๐๐ช๐ฏ๐ฅ-๐๐ฏ๐ง๐ฐ๐ณ๐ฎ๐ฆ๐ฅ ๐๐ฆ๐ฆ๐ฅ๐ฃ๐ข๐ค๐ฌ
๐๐ฉ๐ฃ๐ฅ ๐ช๐ผ๐ฟ๐ธ๐๐ต๐ผ๐ฝ
๐๐-๐๐ข๐ค๐: ๐๐ด๐ฆ๐ณ-๐ฑ๐ณ๐ฆ๐ฅ๐ช๐ค๐ต๐ข๐ฃ๐ญ๐ฆ ๐๐ช๐ฏ๐ฆ-๐จ๐ณ๐ข๐ช๐ฏ๐ฆ๐ฅ ๐๐ข๐ค๐ฆ ๐๐ฉ๐ข๐ฑ๐ฆ ๐๐ฅ๐ช๐ต๐ช๐ฏ๐จ
๐๐
๐๐ช๐ง๐ง๐๐บ๐ฆ๐๐บ๐ฏ: ๐๐ด๐ฆ๐ณ-๐ด๐ฑ๐ฆ๐ค๐ช๐ง๐ช๐ค ๐๐ถ๐ฃ๐ต๐ญ๐ฆ ๐๐บ๐ฆ ๐๐ฐ๐ท๐ฆ๐ฎ๐ฆ๐ฏ๐ต ๐๐บ๐ฏ๐ต๐ฉ๐ฆ๐ด๐ช๐ด ๐๐ด๐ช๐ฏ๐จ ๐๐ช๐ง๐ง๐ถ๐ด๐ช๐ฐ๐ฏ ๐๐ฐ๐ฅ๐ฆ๐ญ๐ด
Congratulations to the team!
For preprints and updates, feel free to visit our website: https://www.collaborative-ai.org/
Satellite Earth-observation (EO) time series in the optical and microwave ranges of the electromagnetic spectrum are often irregular due to orbital patterns and cloud obstruction. Compositing addresses these issues but loses information with respect to vegetation phenology, which is critical for many downstream tasks. Instead, we present TESSERA, a pixel-wise foundation model for multi-modal (Sentinel-1/2) EO time series that learns robust, label-efficient embeddings. During model training, TESSERA uses Barlow Twins and sparse random temporal sampling to enforce invariance to the selection of valid observations. We employ two key regularizers: global shuffling to decorrelate spatial neighborhoods and mix-based regulation to improve invariance under extreme sparsity. We find that for diverse classification, segmentation, and regression tasks, TESSERA embeddings deliver state-of-the-art accuracy with high label efficiency, often requiring only a small task head and minimal computation. To democratize access, adhere to FAIR principles, and simplify use, we release global, annual, 10m, pixel-wise int8 embeddings together with open weights/code and lightweight adaptation heads, thus providing practical tooling for large-scale retrieval and inference at planetary scale. The model training/inference code, downstream task code, and pre-generated embeddings can be accessed at https://github.com/ucam-eo
Garrett Lynch IRL
Josรฉ Oramas
collaborativeai
lalmei
Anil Madhavapeddy