dlib C++ library full changelog

New Features:
Object Detection Tools:
· Added another image scanning tool similar to scan_image_pyramid. This is the new scan_image_boxes object. It allows a user to easily specify an arbitrary set of object boxes which should be checked by an object detector rather than scanning a fixed sized window over the image as is done by the scan_image_pyramid tool. This allows more flexible scanning strategies. For example, it is now possible to use the selective search method implemented by the new find_candidate_object_locations() routine.
· Added the binned_vector_feature_image.
· Upgraded the object_detector so that you can use the adjust_threshold argument for all versions of the operator() method.
· Added remove_unobtainable_rectangles()

Hashing Tools:
· Added a set of new locality sensitive hashing functions meant for use with larger vectors and higher bit sizes than the current LSH tools.
· These are the new hash_similar_angles_xxx objects.

New Stuff:
Image Processing:
· Added the option to make the features generated by the poly_image
· rotationally invariant.
· Added a set of routines for warping, scaling, and resizing images.
· See the new quot;Scaling and Rotatingquot; section of the image processing
· documentation for details.
· Added the heatmap() routine for converting an image into a heatmap.
· Machine Learning
· Updated the sequence labeling trainer to allow the user to set different
· loss values for different labels.
· Added the rls object. It is an implementation of the linear recursive
· least squares algorithm.
· Added the get_option() routines which slightly simplify option parsing
· from the command line and config files.
· Added the 128bit version of Murmur hash.
· Added the rls_filter and kalman_filter objects. These are tools for
· performing Kalman filtering and recursive least squares filtering.
· Added the circular_buffer object.
Non-Backwards Compatible Changes:
· The poly_image generates slightly differen

New Stuff:
· Added tools for timing blocks of code
· Machine Learning
· Added a set of tools for learning to solve the assignment problem.
· See the structural_assignment_trainer and its associated example
· program for an introduction.
· Added random projection based locality sensitive hashing tools.
· Added tools to simplify the creation of scan_image_pyramid objects.
· See the object_detector_ex.cpp example program for details.
· Image Processing
· Added sum_filter() and spatially_filter_image_separable_down()
· New feature extractors: poly_image, nearest_neighbor_feature_image, and
· fine_hog_image

Non-Backwards Compatible Changes:
· Changed the serialization format for rand objects.
· Changed the order of arguments for the sequence_labeler's constructor.
· Object Detection Changes
· Some parts of the object detection tools have been refactored. In particular,
· the interfaces of the scan_image_pyramid and structural_object_detection_trainer
· have been changed slightly to improve

New Stuff:
· Machine Learning
· Added the histogram intersection kernel for sparse and dense vectors.
· Added a set of tools to allow a user to easily learn to do sequence
· labeling using dlib's structural SVM implementation. See the new
· sequence_labeler object and its associated example program for an
· introduction.

Image processing:
· Added segment_image()
· Added randomly_color_image()
· Added the border_enumerator
· Added the disjoint_subsets object, it is an implementation of the
· union-find algorithm/disjoint-set data structure.
· Added new matrix routines: conv(), conv_same(), conv_valid(), xcorr(),
· xcorr_same(), xcorr_valid(), and flip().
Non-Backwards Compatible Changes:
· Changed find_max_factor_graph_viterbi() so you can use run-time
· defined order and num_states parameters.

Bug fixes:
· The last dlib release added a max_iterations parameter to the
· svm_c_linear_trainer and svm_c_ekm_trainer objects.

New Stuff:
Two new routines for performing MAP inference in factor graphs:
· For chain-structured graphs: find_max_factor_graph_viterbi()
· For general graphs: find_max_factor_graph_nmplp()

· Image Processing
· Added more tools for creating image pyramids. See pyramid_down_5_4,
· pyramid_down_4_3, and pyramid_down_3_2.
· Added more image filtering and morphology functions.

Added a set of tools for creating sliding window classifiers:
· Added the scan_image() routine. It is a tool for sliding a set of
· rectangles over an image space and finding the locations where the sum
· of pixels in the rectangles exceeds a threshold. Also added
· scan_image_pyramid, which is a tool for running scan_image() over an
· image pyramid.
· Added the structural_object_detection_trainer. This is a tool which
· formulates the sliding window classifier learning problem as an
· instance of structural SVM learning.
· Added a variety of supporting tools and two object detection example
· programs.

New Stuff:
· Added the check_sub_option() method to the command line parser check
· object.
· Added match_endings to the dir_nav utils.
· Added a set_current_dir() function.
· Added the distance_to_rect_edge() routine.
· Added support for user drawn rectangle overlays and selectable overlays
· to the image_display widget.

Bug fixes:
· Fixed a bug in the image_display widget. If you switched it between
· images of a different size while any kind of zoom was in effect
· it could cause a segmentation fault.