trt-yolo-app detection errors

Accelerated Computing Intelligent Video Analytics DeepStream SDK
1

Hello,

I trained a custom YOLO v3 engine using darknet, and then tried using the trained weights and cfg file with trt-yolo-app (on a Tesla P100).
Although objects were detected, the bounding boxes were not accurate. Sometimes, the bounding box co-ordinates were negative numbers.

I ran the same images through darknet for inference, and got good results.

I would have expected the behavior to be identical. Has anyone else encountered this problem?

Thanks.

2

Hi,

Is there any update in the output semantics meaning?
If yes, please update the parser of bbox here:
[url]https://github.com/NVIDIA-AI-IOT/deepstream_reference_apps/blob/bd5abf33f76b9b051c6d724e1f3aaecb4f1c989c/sources/lib/trt_utils.cpp#L186[/url]

Thanks.

3

Hi,

There’s nothing different in the semantics. It’s just detection of objects with YOLO. Instead of the standard set of 80 objects, I used a different set of 10 or so objects. Training and inference in Darknet worked fine. However, doing this on trt-yolo-app didn’t give the same results.
I would have expected that trt-yolo-app behaves the same way as Darknet, except that the Darknet framework need not be installed.

Thanks.

4

I have the same problem as prashanth.bhat when running https://github.com/NVIDIA-AI-IOT/deepstream_reference_apps/tree/master/sources/samples/objectDetector_YoloV3 with streams in sample of DeepStreamSDK3.

5

Hi,

Have you updated the model information:
[url]https://github.com/NVIDIA-AI-IOT/deepstream_reference_apps/blob/bd5abf33f76b9b051c6d724e1f3aaecb4f1c989c/config/yolov3.txt[/url]

Thanks

6

No. I haven’t. Can you tell me, what I need to update?

7

If there’s no change in the network arch but if only the number of output classes have changed, the trt-yolo-app should work fine. Whats the difference in results you are seeing ?

8

Please modify the the output buffer parser function according to the changes you have made in your network - https://github.com/NVIDIA-AI-IOT/deepstream_reference_apps/blob/master/sources/samples/objectDetector_YoloV3/nvdsinfer_custom_impl_YoloV3/nvdsparsebbox_YoloV3.cpp#L200

9

I am also getting the same issue, the only change is number of classes. I am using tiny yolo v3

10

Did you got any solution ?