Add input_grayscale_image_pyramid, issue #354#1761
Conversation
|
Thanks for the PR. The code looks correct. You should use spaces instead of tabs though. And if you could figure out how to reduce the code duplication that would be good too, like making a base class that contains most of the code and is inherited from. |
|
Sorry about the tabs, I've fixed it already. |
|
Sorry for the late reply. Yeah, that sounds fine. Don't use any virtual functions, you can just publicly inherit from the base class to get all the methods. As for I wouldn't mention this inheritance in the documentation at all. It's an implementation detail users don't need to know about. |
|
With this commit I've create a base clase |
|
And now I've refactored to_tensor method, to move repeated code to base class. |
|
Some benchmarking I've made:
It is around 11% faster. |
davisking
left a comment
davisking
left a comment
There was a problem hiding this comment.
Looks good, just a few minor things and it's ready to merge.
|
Sweet, looks good. Thanks for the PR :) |
* Fixed compiler warnings * Include the Intel MKL's iomp dll in the output folder to reduce confusino for windows users. * Fixed build error in newer clang on OpenBSD. * Fixed constness for lapack functions (davisking#1737) * disable annoying warning * Fixed global_function_search's initialization being wrong if explicitly given an empty list of initial function evaluations. * Suppress compiler warnings * Make things work in visual studio. * fix some pedantic warnings (davisking#1756) * fix some pedantic warnings * remove unneeded assert * more pedantic silencing (davisking#1763) * prevent GCC from complaining about this unused parameter * Even more warning silencing (davisking#1766) These warnings occurred when building the semantic segmentation examples * iEnsures DLIB_FALLTHROUGH macro is only set for GCC>=7 (davisking#1770) * Feature/upgrade libjpeg (davisking#1769) * Upgrades dlib's included libjpeg to version 8d * Overloads load_jpeg to read from memory buffer * Removes "__inline__" define in jconfig, broke VC build * Changes buffer size type to size_t * Adds a comprehensive error message when jpeg loading fails. * Disable use of non-memory based backing store in libjpeg. This fixes libjpeg not being able to open some types of jpeg file. * Stop building parts of libjpeg we don't need. * Add input_grayscale_image_pyramid, issue davisking#354 (davisking#1761) Add input_grayscale_image_pyramid * Added methods for getting keyboard and mouse clicks to image_window's pyhton API. * Fixed pytest broken dependencies * Fix python setup warnings * Revert "Fixed pytest broken dependencies" Apparently pytest is still sort of busted. This reverts commit 5e63d01. * Fix setting a point's y coordinate changes x instead (Python bindings) (davisking#1795) * Add point assignment test * Fix setting points y coordinate changes x instead (issue davisking#1794) * Push all include and link options needed for dlib to pkg-config. We do this by getting them from the same list cmake uses. * Fixed incorrect return type * Fixed grammar in comments * Added missing include * fixed typo in docs * fix mismatch between documentation and implementation (davisking#1835) * Fixed cmake warning * fixing grammar * Fix the CMake BUILDING_PYTHON_IN_MSVC variable not getting picked up where it should. * pybind11: cmake: ignore the check between host-python and cross-compiler (davisking#1848) When dlib is compiling, cmake will compare python architecture and target architecture. So in cross-compiling case, it is irrevelant because host and target architecture often differs. The main problem come from checking python architecture on host and not on target. Here is an error when compiling dlib from x86_64 to arm 32-bit target : ``` Python config failure: Python is 64-bit, chosen compiler is 32-bit ``` So : - Skipping the comparation when cross-compiling is enabled. Signed-off-by: Romain Naour <romain.naour@smile.fr> Signed-off-by: Alexandre PAYEN <alexandre.payen@smile.fr> * Const-correct a LAPACK declaration and add aarch64 as a 64-bit architecture (davisking#1859) * Added aarch64 to list of 64-bit architechtures * Const-corrected declaration of ssyevr * Fix davisking#1849 by calling device_global_buffer() unconditionally (davisking#1862) * Hold on to the CUDA buffer - second try see: davisking#1855 (comment) * Fix davisking#1849 by calling device_global_buffer() unconditionally * Simplified the device_global_buffer() code and API. * don't cast away constness (davisking#1865) * dpoint mutates x-coord in y-property (see davisking#1794) (davisking#1866) * add loss_mean_squared_per_channel (davisking#1863) add loss_mean_squared_per_channel_and_pixel * Clear truth_idxs between samples (davisking#1870) * Clear truth_idxs between samples * Move truth_idxs inside loop body after all * Push to truth_idxs even when the box can't be detected; improve formatting * Add an option to force static runtime (davisking#1847) * dos2unix tell_visual_studio_to_use_static_runtime.cmake * Add an option to force static runtime
Add input_grayscale_image_pyramid
2 participants
Issue #354
I'm doing some test with it, I'll be uploading a performance comparison tomorrow using it in the mmod (the training takes several hours).
In the meantime I wanted to create the pull request, to be sure it's a good contribution, and to start reviewing the code.