Tools - shivamvats/notes GitHub Wiki

• 3DSegmentation: 3D-Box via Segment Anything + VoxelNext.
• Detectron: Facebook's library of deep learning models for perception.
• PyTorch3D: Pytorch library for 3D data, like meshes.
• MDETR: Prompt-based open-world object detection with bounding boxes and segmentation

1. gorilla: Large Language Model Connected with Massive APIs

1. ikdb: Kris Hauser's data-driven IK implementation. Trains a database of globally optimal IK solutions. At test time, it does k-NN and locally optimizes the solution.
2. force_control: Yifan Hou's implementation
3. dmpbbo : Implementations of DMP and black-box optimization.
4. JuliaPOMDP : Interface to write and solve MDPs and POMDPs.
5. Foam : tool to approximate robot URDF using spheres

1. pinocchio : Fast rigid body dynamics and their analytical derivatives.
2. ModernRobotics: Implements all the important manipulation operations.
3. roboticstoolbox : Implements robot kinematics and dynamics in python.

1. Brax : Massively parallel and differentiable rigid body simulator.
2. RaiSim : Fast, C++ based with Python bindings. Actively being developed.
3. Titan : GPU accelerated, C++ based and recently developed.
4. odio_urdf: Procedurally generate urdf files in Python.
5. robosuite:A Modular Simulation Framework and Benchmark for Robot Learning
6. ikea-furniture: Simulation for IKEA furniture assembly
7. google-scanned-objects: 3D scans of objects
8. point-e: Generate 3D models from text prompts

1. reinforcement-learning : Policy Iteration and Value Iteration in Python
2. ray : Framework to build and run distributed applications. Comes packaged with a RL library and a hyperparameter tuning library.
3. flux : Julia library
4. pomegranate: Provides fast and flexible probabilistic models and enables easy composition of distributions, etc.
5. metric-learn : sklearn package for learning a metric from features and labels.
6. bolero : Python library for robot behaviour learning. Provides implementations of DMP, REPS, blackbox optimization and contextual REPS.
7. Gaussian Processes: GPy , GPyTorch_, George (for bigger datasets), gp_extra (non-stationary kernels)
8. pytorch-lightning_ : Leaner pytorch
9. Hyper-parameter Optimization: ray-tune, tune-sklearn
10. PyTorch3D: Pytorch library for 3D data, like meshes.
11. distil : Active learning for deep learning
12. MinkowskiEngine : DL library for sparse tensors.
13. umap : Dimensionality reduction

1. blackbox : Parallel optimization of expensive black-box functions
2. ax : Command-line based hyper-parameter optimization.
3. Decision-Tree for Optimization Tool : has details about and links to many other solvers.
4. Baron : Solver for problems with polynomial constraints

1. tetrad : Software package for causal discovery.

1. Linear Algebra: xtensor

1. Utility: autolab_core
2. Logging: sacred, coloredlogs
3. Experiments: Pandas are great at handling experiment data.
4. Parallel Processing:

• redis-queue : Queuing jobs and processing them in the background with workers.

5. reloading: Hot swap code

1. Log Viewer: lnav
2. journald: Linux daemon to store and view system-wide logs.
3. Syncthing: Sync directories over ssh persistently
4. taskset: taskset --clu-list 0-16 <program> specifies cores to run the program on: cpu affinity.

1. pyclips : Python bindings to the CLIPS rule based programming software.
2. godbolt : Online compiler explorer.
3. ox-hugo : Export org-mode files to html. Also see how Kethro-Kuan does it.
4. manim : Math visualizations.

1. Blender
2. Solidworks
3. Library of CAD models
4. Objaverse: 1 million annotated 3D objects

1. taskjuggler : prepare Gantt charts; integration with org-mode; command: tj3