# How to avoid groupRectangles, boundingboxes and get only class probabilty map

**URL:** <https://forums.developer.nvidia.com/t/how-to-avoid-grouprectangles-boundingboxes-and-get-only-class-probabilty-map/49018>\
**Category:** Jetson TX1\
**Created:** [April 4, 2017, 7:45pm UTC](https://forums.developer.nvidia.com/t/how-to-avoid-grouprectangles-boundingboxes-and-get-only-class-probabilty-map/49018 "2017-04-04T19:45:45Z")\
**Posts on this page:** 6\
**Page:** 1

<div class="post-metadata">

**Author:** ![dodo91](https://developer.download.nvidia.com/images/forums/profile-default-devtalk-84.png) [@dodo91](https://forums.developer.nvidia.com/u/dodo91)\
**Post date:** [April 4, 2017, 7:45pm UTC](https://forums.developer.nvidia.com/t/how-to-avoid-grouprectangles-boundingboxes-and-get-only-class-probabilty-map/49018/1 "2017-04-04T19:45:45Z")

</div>

What should I do, if i want to get that output before the clustering Python Laye i dont want to get back boundigboxes. (How to get “real-valued multi-dimensional array that can be overlaid on the image, much like the DetectNet input labels described above”, how to avoid Python Layer that run groupRectangles funciton?)  
Thanks

---

<div class="post-metadata">

**Author:** ![AastaLLL](https://sea2.discourse-cdn.com/nvidia/user_avatar/forums.developer.nvidia.com/aastalll/32/14043_2.png) [@AastaLLL](https://forums.developer.nvidia.com/u/AastaLLL)\
**Post date:** [April 5, 2017, 7:09am UTC](https://forums.developer.nvidia.com/t/how-to-avoid-grouprectangles-boundingboxes-and-get-only-class-probabilty-map/49018/2 "2017-04-05T07:09:18Z")

</div>

Hi,

Thanks for your question.

May I know which model and inference framework you used?  
If you using DetectNet with tensorRT, please remove last python layer and use bboxes and coverage as output.

---

<div class="post-metadata">

**Author:** ![dodo91](https://developer.download.nvidia.com/images/forums/profile-default-devtalk-84.png) [@dodo91](https://forums.developer.nvidia.com/u/dodo91)\
**Post date:** [April 5, 2017, 11:57am UTC](https://forums.developer.nvidia.com/t/how-to-avoid-grouprectangles-boundingboxes-and-get-only-class-probabilty-map/49018/3 "2017-04-05T11:57:35Z")

</div>

Thanks for the answer, and the coverage output is a value for each grid (between 0-1) and tell the the likelihood of the class present or not in the grid or in bbox? If it tell in the bbox, how can i get the likelihood of the class present in the grid?

---

<div class="post-metadata">

**Author:** ![AastaLLL](https://sea2.discourse-cdn.com/nvidia/user_avatar/forums.developer.nvidia.com/aastalll/32/14043_2.png) [@AastaLLL](https://forums.developer.nvidia.com/u/AastaLLL)\
**Post date:** [April 7, 2017, 6:12am UTC](https://forums.developer.nvidia.com/t/how-to-avoid-grouprectangles-boundingboxes-and-get-only-class-probabilty-map/49018/4 "2017-04-07T06:12:38Z")

</div>

Hi,

Thanks for your feedback.

Bbox contains detail location information while coverage is the object probability.  
If you are interested in likelihood, please use coverage.

---

<div class="post-metadata">

**Author:** ![dodo91](https://developer.download.nvidia.com/images/forums/profile-default-devtalk-84.png) [@dodo91](https://forums.developer.nvidia.com/u/dodo91)\
**Post date:** [April 12, 2017, 8:10am UTC](https://forums.developer.nvidia.com/t/how-to-avoid-grouprectangles-boundingboxes-and-get-only-class-probabilty-map/49018/5 "2017-04-12T08:10:56Z")

</div>

Thanks for your answer!

---

<div class="post-metadata">

**Author:** ![TomNVIDIA](https://sea2.discourse-cdn.com/nvidia/user_avatar/forums.developer.nvidia.com/tomnvidia/32/14181_2.png) [@TomNVIDIA](https://forums.developer.nvidia.com/u/TomNVIDIA)\
**Post date:** [October 18, 2021, 6:59pm UTC](https://forums.developer.nvidia.com/t/how-to-avoid-grouprectangles-boundingboxes-and-get-only-class-probabilty-map/49018/6 "2021-10-18T18:59:43Z")

</div>

Closed on October 18, 2021, 6:59pm UTC