tag:blogger.com,1999:blog-6061887630060661987.post6651459553903347452..comments2024-03-13T13:54:35.259-04:00Comments on dlib C++ Library: Dlib 18.6 released: Make your own object detector!Davis Kinghttp://www.blogger.com/profile/16577392965630448489noreply@blogger.comBlogger252125tag:blogger.com,1999:blog-6061887630060661987.post-34700377012973909272019-02-24T18:45:38.864-05:002019-02-24T18:45:38.864-05:00There are a number of face detection examples in d...There are a number of face detection examples in dlib's examples folder. They don't require training your own model.Davis Kinghttps://www.blogger.com/profile/16577392965630448489noreply@blogger.comtag:blogger.com,1999:blog-6061887630060661987.post-50915503421114902482019-02-24T05:24:36.142-05:002019-02-24T05:24:36.142-05:00Is the simple face_detector.svm file somewhere dow...Is the simple face_detector.svm file somewhere downloadable ? I am doing some performance tests on a mobile and I want to avoid setting up a training on OS-X when I see beforehand that performance is too slow. <br />Thanks ChrisAnonymoushttps://www.blogger.com/profile/05757629158617809492noreply@blogger.comtag:blogger.com,1999:blog-6061887630060661987.post-15943559343757312982019-01-11T23:49:07.640-05:002019-01-11T23:49:07.640-05:00Hi. I want to know can I implement dlid to detect ...Hi. I want to know can I implement dlid to detect another shape not only face shape, but wrinkles on the face?<br />Thank you for your answerAnonymoushttps://www.blogger.com/profile/05924672442200183084noreply@blogger.comtag:blogger.com,1999:blog-6061887630060661987.post-29657072678865862382018-10-22T11:23:37.203-04:002018-10-22T11:23:37.203-04:00Hi,
I'm quite new to this, but apparently, my...Hi, <br />I'm quite new to this, but apparently, my code seems to lack a header file for including the Python library. Please provide suggestions to include the same.I've already checked the version of python installed and i've already verified if the python2.7 installed is working properly.<br />The error was encountered at 83% while building the CXX object file. I've attatched a copy of the error below : <br /><br />/home/home-pc/fd_fr/dlib/dlib/external/pybind11/include/pybind11/detail/common.h:111:10: fatal error: Python.h: No such file or directory<br /> #include <br /> ^~~~~~~~~~<br />compilation terminated.<br />CMakeFiles/dlib_python.dir/build.make:62: recipe for target 'CMakeFiles/dlib_python.dir/src/dlib.cpp.o' failed<br />make[2]: *** [CMakeFiles/dlib_python.dir/src/dlib.cpp.o] Error 1<br />CMakeFiles/Makefile2:67: recipe for target 'CMakeFiles/dlib_python.dir/all' failed<br />make[1]: *** [CMakeFiles/dlib_python.dir/all] Error 2<br />Makefile:83: recipe for target 'all' failed<br />make: *** [all] Error 2Anonymoushttps://www.blogger.com/profile/10153470939692548936noreply@blogger.comtag:blogger.com,1999:blog-6061887630060661987.post-35936723015124054272018-07-04T20:29:02.121-04:002018-07-04T20:29:02.121-04:00The details are in this post: http://blog.dlib.net...The details are in this post: http://blog.dlib.net/2017/02/high-quality-face-recognition-with-deep.htmlDavis Kinghttps://www.blogger.com/profile/16577392965630448489noreply@blogger.comtag:blogger.com,1999:blog-6061887630060661987.post-33635064617884473232018-07-04T12:38:33.467-04:002018-07-04T12:38:33.467-04:00Would you please tell me how many faces would be e...Would you please tell me how many faces would be enough to train the "dlib_face_recognition_resnet_model_v1" from scratch? or how many images did you use to train it?<br />ThanksAnonymoushttps://www.blogger.com/profile/02464043860340726971noreply@blogger.comtag:blogger.com,1999:blog-6061887630060661987.post-26804229502327601262018-07-04T06:45:07.189-04:002018-07-04T06:45:07.189-04:00It's not a public dataset.It's not a public dataset.Davis Kinghttps://www.blogger.com/profile/16577392965630448489noreply@blogger.comtag:blogger.com,1999:blog-6061887630060661987.post-87495990401455780282018-07-03T19:40:35.582-04:002018-07-03T19:40:35.582-04:00Would you please let me know what data-set you use...Would you please let me know what data-set you used to train the "dlib_face_recognition_resnet_model_v1"?<br />ThanksAnonymoushttps://www.blogger.com/profile/02464043860340726971noreply@blogger.comtag:blogger.com,1999:blog-6061887630060661987.post-30611789808699894222018-07-03T19:38:57.136-04:002018-07-03T19:38:57.136-04:00This comment has been removed by the author.Anonymoushttps://www.blogger.com/profile/02464043860340726971noreply@blogger.comtag:blogger.com,1999:blog-6061887630060661987.post-13442865660999441782018-03-06T22:07:38.531-05:002018-03-06T22:07:38.531-05:00Good to know!
Really looking forward to AVX-512 ve...Good to know!<br />Really looking forward to AVX-512 version of dlib's HOG face detector!<br />Thanks for the continuous update and maintenance to dlib ;)Anonymoushttps://www.blogger.com/profile/07913588115407332454noreply@blogger.comtag:blogger.com,1999:blog-6061887630060661987.post-45539499427031523052018-03-06T06:48:28.130-05:002018-03-06T06:48:28.130-05:00No, I'm not sure there is any benefit. At som...No, I'm not sure there is any benefit. At some point AVX-512 will be used though.Davis Kinghttps://www.blogger.com/profile/16577392965630448489noreply@blogger.comtag:blogger.com,1999:blog-6061887630060661987.post-39200795902244932742018-03-06T04:15:53.773-05:002018-03-06T04:15:53.773-05:00Dear Davis:
I enabled AVX instructions and there i...Dear Davis:<br />I enabled AVX instructions and there is huge boost in performance! Thanks!<br />Any plan in implementing AVX2 to dlib's HOG face detector, and will it be even faster?Anonymoushttps://www.blogger.com/profile/07913588115407332454noreply@blogger.comtag:blogger.com,1999:blog-6061887630060661987.post-83739912607937896572018-01-28T21:40:40.701-05:002018-01-28T21:40:40.701-05:00Make sure you have AVX instructions enabled. That ...Make sure you have AVX instructions enabled. That makes a big difference.Davis Kinghttps://www.blogger.com/profile/16577392965630448489noreply@blogger.comtag:blogger.com,1999:blog-6061887630060661987.post-47680386941828646382018-01-28T21:38:21.203-05:002018-01-28T21:38:21.203-05:00Hi Davis:
I used Intel's MKL to speed up face ...Hi Davis:<br />I used Intel's MKL to speed up face recognition as you suggested and gained good speed enhancements, however dlib's HOG face detection seems to slow down using MKL's ICC++ compared to vs++ original compiler. I analyzed via Intel's VTUNE and the bottleneck seems to be in dlib\image_transforms\spatial_filtering.h when applying the row & col filters. Is there any insight you might suggest to speed performance, as I use up 90 ticks (about 11 FPS) for 960*540 resolution in dlib's HOG face detection ?Anonymoushttps://www.blogger.com/profile/07913588115407332454noreply@blogger.comtag:blogger.com,1999:blog-6061887630060661987.post-66149738922870278472017-12-04T06:23:06.788-05:002017-12-04T06:23:06.788-05:00Hi Davis, I have been working from last few months...Hi Davis, I have been working from last few months to use dlib for face detection in video. Due to requirements to detect side pose for a commercial app, I have trained my own detector. It has been working on 70% of the side pose frames. But the problem is for few frames, it doesn't detect the face; even though it does detect for the very next frame. Even though difference in these two adjacent frames is minimal. Can you please give any pointers on how to debug the problem? Anonymoushttps://www.blogger.com/profile/15477540456586620112noreply@blogger.comtag:blogger.com,1999:blog-6061887630060661987.post-68690524561248111642017-08-06T13:13:56.557-04:002017-08-06T13:13:56.557-04:00Is there any method of implementing Genetic algori...Is there any method of implementing Genetic algorithms on dlib? Any bio inspired algorithms?Subhenduhttps://www.blogger.com/profile/02522003350134973626noreply@blogger.comtag:blogger.com,1999:blog-6061887630060661987.post-46400053700777335162017-08-06T13:10:41.475-04:002017-08-06T13:10:41.475-04:00I am trying to use both dlib and opencv together i...I am trying to use both dlib and opencv together in my python programs. <br />Can you please tell me how to grab video from webcam.?<br /><br />Is there any method for detecting pedestrians from car camera? I am looking for a real time solution. <br /><br />Thanks for this beautiful librarySubhenduhttps://www.blogger.com/profile/02522003350134973626noreply@blogger.comtag:blogger.com,1999:blog-6061887630060661987.post-81014281388132956882017-06-13T06:16:33.946-04:002017-06-13T06:16:33.946-04:00Hi Davis,
I'm new to dlib, I notice that dlib...Hi Davis,<br /><br />I'm new to dlib, I notice that dlib supports Deep Learning, but does it support using Deep Learning for Pedestrian Detection? If yes, could you advise on how to do this? Thanks a lot for your help and for a great dlib!<br /><br />Best,<br />phivuAnonymoushttps://www.blogger.com/profile/04687636171495413543noreply@blogger.comtag:blogger.com,1999:blog-6061887630060661987.post-52135219593324684452017-06-09T20:52:46.724-04:002017-06-09T20:52:46.724-04:00I don't know what to tell you. Look at the ne...I don't know what to tell you. Look at the network architecture and see what it does. It's running convolutionally over the image. There is nothing surprising about it's speed. Obviously if you used region proposals and recomputed things for each proposal then it would be terribly slow. But that's a silly thing to do. I realize lots of people do it, but it's not super clear to me why when it's easy to run an entire image pyramid through a CNN in a convolutional mode. That's super fast and covers all scales and box positions. Davis Kinghttps://www.blogger.com/profile/16577392965630448489noreply@blogger.comtag:blogger.com,1999:blog-6061887630060661987.post-46452669997372621602017-06-09T20:30:38.486-04:002017-06-09T20:30:38.486-04:00Yes but region proposal methods like faster rcnn o...Yes but region proposal methods like faster rcnn only run the detection (for example a cnn) on the proposed regions right? where as mmod runs it on all possible regions? Dont you have millions of regions and running a forward pass on millions of regions is very computationally expensive?Anonymoushttps://www.blogger.com/profile/07648619871888054997noreply@blogger.comtag:blogger.com,1999:blog-6061887630060661987.post-50941776722464057292017-06-09T20:09:19.016-04:002017-06-09T20:09:19.016-04:00Computationally, it's just a matter of running...Computationally, it's just a matter of running the detector on an image. If that were too expensive then the whole thing would be pointless because the detector would be too slow to be useful once learned.Davis Kinghttps://www.blogger.com/profile/16577392965630448489noreply@blogger.comtag:blogger.com,1999:blog-6061887630060661987.post-47112563760732201782017-06-09T19:48:49.596-04:002017-06-09T19:48:49.596-04:00But isnt it still expensive to consider all window...But isnt it still expensive to consider all windows? You consider one window size on every level of the pyramid right (in the new model with the cnn)?Anonymoushttps://www.blogger.com/profile/07648619871888054997noreply@blogger.comtag:blogger.com,1999:blog-6061887630060661987.post-58682119060590491702017-06-09T19:26:36.869-04:002017-06-09T19:26:36.869-04:00It's not easy to figure out how to consider al...It's not easy to figure out how to consider all windows efficiently and in a way that makes sense. That's the central innovation of the paper.Davis Kinghttps://www.blogger.com/profile/16577392965630448489noreply@blogger.comtag:blogger.com,1999:blog-6061887630060661987.post-20523961842941719292017-06-09T16:43:40.340-04:002017-06-09T16:43:40.340-04:00Hi David,
I read in your paper that compared to ot...Hi David,<br />I read in your paper that compared to other detection methods, you are considering all windows (with a given aspect ratio). My question is, why are other methods only considering a subset if you it is possible to consider all windows like you do. And isn't it very expensive to compute \phi for every window?Anonymoushttps://www.blogger.com/profile/07648619871888054997noreply@blogger.comtag:blogger.com,1999:blog-6061887630060661987.post-50939862547788496232017-06-04T15:34:25.043-04:002017-06-04T15:34:25.043-04:00Make sure you aren't upsampling the images in ...Make sure you aren't upsampling the images in the code a bunch. But it does take a good amount of RAM.Davis Kinghttps://www.blogger.com/profile/16577392965630448489noreply@blogger.com