Version 8.2, part of Release 2018b, includes the following enhancements:

  • Video Labeler App: Interactive and semi-automatic labeling of ground truth data in a video, image sequence, or custom data source
  • Lidar Segmentation: Segment ground points from organized 3-D lidar data and organize point clouds into clusters
  • Point Cloud Registration: Align 3-D point clouds using coherent point drift (CPD) registration
  • MSAC Fitting: Find a polynomial that best fits noisy data using the M-estimator sample consensus (MSAC)
  • Faster R-CNN Enhancements: Train Faster R-CNN object detectors using DAG networks such as ResNet-50 and Inception-v3

See the Release Notes for details.

Version 8.1, part of Release 2018a, includes the following enhancements:

  • Lidar Segmentation: Segment lidar point clouds using Euclidean distance
  • Lidar Registration: Register multiple lidar point clouds using normal distributions transform (NDT)
  • Image Labeler App: Mark foreground and background for pixel labeling
  • Fisheye Calibration: Interactively calibrate fisheye lenses using the Camera Calibrator app
  • Stereo Baseline Estimation: Estimate baseline of a stereo camera with known intrinsic parameters

See the Release Notes for details.

Version 8.0, part of Release 2017b, includes the following enhancements:

  • Semantic Segmentation Using Deep Learning: Classify pixel regions in images, evaluate, and visualize segmentation results
  • Image Labeling App: Interactively label individual pixels for semantic segmentation and label regions using bounding boxes for object detection
  • Fisheye Camera Calibration: Calibrate fisheye cameras to estimate intrinsic camera parameters
  • KAZE Features: Detect and extract KAZE features for object recognition or image registration workflows

See the Release Notes for details.

Version 7.3, part of Release 2017a, includes the following enhancements:

  • Deep Learning for Object Detection: Detect objects using Fast R-CNN and Faster R-CNN object detectors
  • Object Detection Using ACF: Train object detectors using aggregate channel features
  • Object Detector Evaluation: Evaluate object detector performance, including precision and miss-rate metrics
  • OpenCV Interface: Integrate OpenCV version 3.1.0 projects with MATLAB

See the Release Notes for details.

Version 7.2, part of Release 2016b, includes the following enhancements:

  • Deep Learning for Object Detection: Detect objects using region-based convolution neural networks (R-CNN)
  • Structure from Motion: Estimate the essential matrix and compute camera pose from 3-D to 2-D point correspondences
  • Point Cloud File I/O: Read and write PCD files using Point Cloud File I/O Functions
  • Code Generation for ARM Example: Detect and track faces on a Raspberry Pi 2 target
  • Visual Odometry Example: Estimate camera locations and trajectory from an ordered sequence of images

See the Release Notes for details.

Version 7.1, part of Release 2016a, includes the following enhancements:

  • OCR Trainer App: Train an optical character recognition (OCR) model to recognize a specific set of characters
  • Structure from Motion: Estimate the camera poses and 3-D structure of a scene from multiple images
  • Pedestrian Detection: Locate pedestrians in images and video using aggregate channel features (ACF)
  • Bundle Adjustment: Refine estimated locations of 3-D points and camera poses for the structure from motion (SFM) framework
  • Multiview Triangulation: Triangulate 3-D locations of points matched across multiple images

See the Release Notes for details.