This page is dedicated to software packages and datasets produced by the Thoth team (since 2016).
For older work, see the page of the LEAR team
Datasets
- The AVA Dataset: A video dataset of spatio-temporally localized atomic visual actions.
- The DALY Dataset: A dataset for spatio-temporal action detection.
- Charades-Ego Dataset: A dataset of videos with first and third person views.
Software
- The Cyanure toolbox: A collection of fast solvers for linear models in machine learning.
- The SPAMS toolbox: A collection of fast solvers for sparse estimation and dictionary learning in machine learning.
- CKN in C++/Matlab: A C++ and matlab implementation of convolutional kernel networks.
- Multimodal Transformer: Software package for cross-modal video retrieval.
- Action Recognition: Package for action recognition based on motion-augmented RGB.
- Incremental Learning: Software package for end-to-end incremental image classification.
- First and Third Person Activites: Joint modelling of videos from first and third person perspectives.
- Video Object Segmentation: Software package for learning video object segmentation.
- Loter: A Software Package to Infer Local Ancestry for a Wide Range of Species.
- BlitzNet: A Real-Time Deep Network for Scene Understanding.
- Modl: Dictionary Learning for Massive Matrix Factorization.
- CKN-seq: Convolutional Kernel Networks for Biological Sequence Modeling.
- Context-DA: Context-driven data augmentation for object detection
- Few-shot learning: Ensemble Methods for Few-Shot Classification
- SUR: Selecting from Universal Representations
- RKN: Recurrent Kernel Networks
- GCKN: Graph Convolutional Kernel Networks
- OTK: Optimal Transport Kernel Networks
- GraphiT: Transformer Model for Graphs
- CogSpaces: Multi-study decoding models for task functional MRI data
- Stochs: C++ library with Cython bindings for fast stochastic optimization algorithms for machine learning
- GroupSC: Fully Trainable and Interpretable Non-Local Sparse Models for Image Restoration