103 alexander alekhin
25 dmitry kurtaev
23 Maksim Shabunin
17 li peng
17 Elenagvo
14 Vitaly Tuzov
10 Suleyman Turkmen
8 Tomoaki Teshima
6 sayed adel
5 Laurentberger
5 Vladislav Sovrasov
5 wu zhiwen
4 Pavel Rojtberg
4 rastislav vasilikhin
3 tribta
2 Bhanudutta
2 Fakabbir Amin
2 mattia rizzolo
2 Ryan Fox
2 shinya ishikawa
2 berak
2 dtmoodie
1 akhilesh kumar
1 Alexander Nesterov
1 alexander rybnikov
1 amro
1 Andrey Smorodov
1 Arthur Pastel
1 Cartucho
1 Christof Kaufmann
1 David Geldreich
1 Florian Echtler
1 Hamdi Sahloul
1 Haritha
1 IAGO Suárez
1 igor wodiany
1 Ivan pozdeev
1 Jacob Macritchie
1 James Perkins
1 jcrist99
1 jiri horner
1 Jonathan Viney
1 juha reunanen
1 kuang fangjun
1 Mikhail Paulyshka
1 Muhammad Abdullah
1 nickola
1 pushkal katara
1 riyuzakii
1 roman cattaneo
1 Shresth Verma
1 wei hao
1 wu, zhiwen
1 Alessandro Faria
1 gdkessler
1 klchang
1 woody.chow
1 wxzs5
1 zhijackchen
1 zhongwuzw

OPENCV 3.4-2017年新年礼物-老树发新芽

18 Vladislav Sovrasov
9 Alexander Alekhin
6 Vitaly Tuzov
6 berak
5 Hamdi Sahloul
5 Maksim Shabunin
4 Pavel Rojtberg
4 Suman Ghosh
3 Laurentberger
3 Leonardo BronS
2 Suleyman Turkmen
1 adam gradzki
1 Anup Parikh
1 dmitry kurtaev
1 egor pugin
1 Leonardo Lontra
1 vladislav samsonov
1 fiammante
1 klchang
1 kurnianggoro
1 kushalvyaskv
1 sukhad-app

I applied this picture of this official website. OpenCV 3.4 grabbed the version of the Christmas Eve, which has a taste for Christmas gifts. In 1999, OPENCV launched from Intel Research Agency to today, almost 18 years. For software, 18 years are almost a century -old length, but OpenCV jumps alive in the memory of computer visual people from generation to generation. With the development of deep learning in the CV industry, OpenCV is getting more and more unbelievable, and countless small fresh meat jumps out. Whether it is Darknet (although it has an OpenCV switch in Makefile Opencv), dlib, is still those deep neural networks that do not make the name, each of which is embarrassed, and the pictures and videos are detailed. In this context, OpenCV 3.4 was released on December 23, 2017.

Let’s take a good look, what is this version.

First, this version integrates a project from the 2017 Google Code of the Summer Camp to separate the background of the picture. Looking at the pictures released from the official website, the effect is general.

Second, OpenCV_DNN continues to expand, adding the Faster R-CNN of He Kaiming’s great god. Of course, everyone generally goes to pilgrimage (https://github.com/rbgirshick/py-faster-rcnn).

Third, OpenCV_DNN supports JavaScript binding. In short, you can run a neural network in the browser. tutorial_dnn_javascript.html). It can indeed run, that is, downloading the trained model at the beginning will take some time, depending on the wall or outside the wall. This is very convenient for many applications.

Fourth, a buddy named Peng Li is amazing and optimizes the MobileNet-SSD network 7 times. This is an era of Chinese. But GitHub couldn’t find more information about this buddy. Someone who knew this magical buddy, please tell Dedman that I just follow him (https://github.com/pengli).

Fifth, this version uses pre -compilation to reduce the initialization time of applications. That is a space to change time.

Sixth, support the Transparent API to load OpenCL, so that some embedded system systems are a gospel using OpenCV.

OPENCV 3.4-2017年新年礼物-老树发新芽

Seventh, I realized the consistency of some cross -platform implementation. I haven’t had time to run the code for this point, so I will not express their opinions for the time being.

Eighth, the cross quotation of some development documents is for children in the summer camp to help. Some examples have been added, such as the calculation of single -stroke matrix, and how to debug the OpenCV code through Intel tools. Of course, the exciting example of the neural network in the browser was also contributed by the children in the summer camp.

Ninth, that is to change the bug, more than 200 bugs fixed, and the patch of 250 people of each passerby merged in.

Finally, thank you for open source developers.



Therefore, after watching this CHANGELOG, I have to say that OpenCV is old. But the veterans have the pride of veterans. Although we will depend on the framework of TensorFlow and TORCH, the requirements for the feature engineering of the picture have weakened, but we will still find that we have a lot of exploration process, and application code, such as calling, such as calling The camera shows all kinds of special effects pictures, which still depends on OpenCV seriously, which reminds me of another open source bull and ffmpeg that is as well as OpenCV. Well, we talk about ffmpeg again when we have time.

You might also enjoy: