Computer Vision Toolbox
Design and test computer vision, 3D vision, and video processing systems
Computer Vision Toolbox™ provides algorithms, functions, and apps for designing and testing computer vision, 3D vision, and video processing systems. You can perform object detection and tracking, as well as feature detection, extraction, and matching. For 3D vision, the toolbox supports single, stereo, and fisheye camera calibration; stereo vision; 3D reconstruction; and lidar and 3D point cloud processing. Computer vision apps automate ground truth labeling and camera calibration workflows.
You can train custom object detectors using deep learning and machine learning algorithms such as YOLO v2, Faster R-CNN, and ACF. For semantic segmentation you can use deep learning algorithms such as SegNet, U-Net, and DeepLab. Pretrained models let you detect faces, pedestrians, and other common objects.
You can accelerate your algorithms by running them on multicore processors and GPUs. Most toolbox algorithms support C/C++ code generation for integrating with existing code, desktop prototyping, and embedded vision system deployment.
Object Detection and Recognition
Frameworks to train, evaluate, and deploy object detectors such as YOLO v2, Faster R-CNN, ACF, and Viola-Jones. Object recognition capability includes bag of visual words and OCR. Pretrained models detect faces, pedestrians, and other common objects.
Segment images and 3D volumes by classifying individual pixels and voxels using networks such as SegNet, FCN, U-Net, and DeepLab v3+.
Ground Truth Labeling
Automate labeling for object detection, semantic segmentation, and scene classification using the Video Labeler and Image Labeler apps.
Lidar and Point Cloud I/O
Read, write, and display point clouds from files, lidar, and RGB-D sensors.
Point Cloud Registration
Register 3D point clouds using Normal-Distributions Transform (NDT), Iterative Closest Point (ICP), and Coherent Point Drift (CPD) algorithms.
Segmentation and Shape Fitting
Segment point clouds into clusters and fit geometric shapes to point clouds. Segment ground plane in lidar data for automated driving and robotics applications.
Single Camera Calibration
Automate checkerboard detection and calibrate pinhole and fisheye cameras using the Camera Calibrator app.
Stereo Camera Calibration
Calibrate a stereo pair to compute depth and reconstruct 3D scenes.
Feature Detection, Extraction, and Matching
Detect, extract, and match interesting features such as blobs, edges, and corners across multiple images.
Feature-Based Image Registration
Match features across multiple images to estimate geometric transforms between images and register image sequences.
Video and Image Labeler
Copy and paste pixel labels; improved pan and zoom; improved frame navigation; line ROI, label attributes, and sublabels added to Image Labeler
Data Augmentation for Object Detectors
Transform image and bounding box
Classify individual pixels in images and 3D volumes using DeepLab v3+ and 3D U-Net networks
Deep Learning Object Detection
Perform faster R-CNN end-to-end training, anchor box estimation, and use multichannel image data
Deep Learning Acceleration
Optimize YOLO v2 and semantic segmentation using MEX acceleration