Research tools for autonomous systems using Python
Author: Sudeep Pillai (spillai@csail.mit.edu)
License: MIT

geometry: General-purpose tools for computing rigid-body transformations. This is a preliminary version that currently deals mostly with SE(3) or 6-DOF (3-DoF Rotational + 3-DoF translation) and some support for Sim(3) motions.

vision: Computer vision package with several tools including camera, tracking, 2d features, 3d features, optical flow, recognition, object proposals, caffe, classifier training, bag-of-words training, geometry, stereo, drawing etc.

utils: Basic tooling that includes attribute dictionaries, database-utils including incremental hdf5 tables, dataset readers [ImageDatasets, StereoDatasets, VelodyneDatasets etc], dataset helpers [KITTI, NYU-RGBD, SUN3D, Tsukuba Stereo, , VaFRIC, UW-RGBD], itertools recipes, timing/profiling tools, io utils [video/image writing, mkdirs, find_files, config parsers, joblib utils, stdout tee-ing, JSON], other misc tools including pretty prints, progressbars, colored prints, counters, accumulators (indexed deques), accumulators with periodic callbacks etc.

externals: ROS/LCM drawing tools, ROS/LCM log readers, Google Project Tango log reader

New (upcoming) modules: mapping, nnet with support for visual odometry, vSLAM, semi-dense reconstruction, and more recent developments in CNN-based object recognition, localization (ROI pooling), and SLAM-aware recognition. See references for more details about specific implementations.

See INSTALL