# Looking for a starting point

**URL:** <https://forum.opencv.org/t/looking-for-a-starting-point/6797>\
**Category:** Python\
**Created:** [December 19, 2021, 5:57pm UTC](https://forum.opencv.org/t/looking-for-a-starting-point/6797 "2021-12-19T17:57:42Z")\
**Posts on this page:** 4\
**Page:** 1

<div class="post-metadata">

**Author:** ![Manuel\_Dahmen](https://sea2.discourse-cdn.com/flex020/user_avatar/forum.opencv.org/manuel_dahmen/32/4052_2.png) [@Manuel\_Dahmen](https://forum.opencv.org/u/Manuel_Dahmen)\
**Post date:** [December 19, 2021, 5:57pm UTC](https://forum.opencv.org/t/looking-for-a-starting-point/6797/1 "2021-12-19T17:57:43Z")

</div>

I’m new to OpenCV and I search for a starting point for training CascadeHaar model for object detection. I have true positive (images with the object), negative (background images without the object) and I’ve just download OpenCV for work either on Windows and command line or Python.

---

<div class="post-metadata">

**Author:** ![berak](https://avatars.discourse-cdn.com/v4/letter/b/85f322/32.png) [@berak](https://forum.opencv.org/u/berak)\
**Post date:** [December 20, 2021, 9:52am UTC](https://forum.opencv.org/t/looking-for-a-starting-point/6797/2 "2021-12-20T09:52:52Z")

</div>

what are you trying to detect ?  
(cascades need _rigid_ objects with reproducable texture)

[the tutorial is here](https://docs.opencv.org/4.x/dc/d88/tutorial_traincascade.html)

you will need the 3.x version of opencv, those tools do no more exist in 4.x

---

<div class="post-metadata">

**Author:** ![Alejandro\_Silvestri](https://sea2.discourse-cdn.com/flex020/user_avatar/forum.opencv.org/alejandro_silvestri/32/357_2.png) [@Alejandro\_Silvestri](https://forum.opencv.org/u/Alejandro_Silvestri)\
**Post date:** [December 20, 2021, 12:46pm UTC](https://forum.opencv.org/t/looking-for-a-starting-point/6797/3 "2021-12-20T12:46:22Z")

</div>

Hi @berak ,

Yes, the reference says “Createsamples and traincascade are disabled since OpenCV 4.0”. Do you know why?

I believe it is because cascades became obsolete since deep learning appeared on scene 10 years ago. Am I right? Is there any other reason you can think of?

---

<div class="post-metadata">

**Author:** ![berak](https://avatars.discourse-cdn.com/v4/letter/b/85f322/32.png) [@berak](https://forum.opencv.org/u/berak)\
**Post date:** [December 20, 2021, 1:53pm UTC](https://forum.opencv.org/t/looking-for-a-starting-point/6797/4 "2021-12-20T13:53:11Z")

</div>

> [@Alejandro\_Silvestri](#):
>
> “Createsamples and traincascade are disabled since OpenCV 4.0”. Do you know why?

the code relied heavily on the removed C-API, and can no more be built

> [@Alejandro\_Silvestri](#):
>
> I believe it is because cascades became obsolete since deep learning appeared on scene 10 years ago

that’s another reason. dnns like yolo are far more robust, and can detect multiple classes at once.