citySemSegFormer deployment config error

• Hardware - Jetson AGX Orin
• Network Type - citySemSegFormer
• TLT Version - N/A, running everything inside nvcr.io/nvidia/deepstream-l4t:6.3-samples, inside it downloaded and built GitHub - NVIDIA-AI-IOT/deepstream_tao_apps: Sample apps to demonstrate how to deploy models trained with TAO on DeepStream. It has TRT version 8.5.2
• Training spec file - N/A
• How to reproduce the issue - try following this CitySemSegformer | NVIDIA NGC

So, essentially I am trying to deploy citySemSegFormer following guide from your site, but on the very last step - when I am trying to test the deployment (typing command ds-tao-segmentation -c configs/segformer_tao/pgie_tlt_config_citysemsegformer.txt -i $DS_SRC_PATH/samples/streams/sample_720p.h264) I’m facing the following error:

(process:1278): GLib-WARNING **: 20:23:28.798: GError set over the top of a previous GError or uninitialized memory.
This indicates a bug in someone's code. You must ensure an error is NULL before it's set.
The overwriting error message was: Key file does not have group ?property?

(process:1278): GLib-WARNING **: 20:23:28.798: GError set over the top of a previous GError or uninitialized memory.
This indicates a bug in someone's code. You must ensure an error is NULL before it's set.
The overwriting error message was: Key file does not have group ?property?

(process:1278): GLib-WARNING **: 20:23:28.798: GError set over the top of a previous GError or uninitialized memory.
This indicates a bug in someone's code. You must ensure an error is NULL before it's set.
The overwriting error message was: Key file does not have group ?property?
No protocol specified
No EGL Display 
nvbufsurftransform: Could not get EGL display connection
Request sink_0 pad from streammux

(ds-tao-segmentation:1278): GLib-GObject-WARNING **: 20:23:28.879: g_object_set_is_valid_property: object class 'GstNvSegVisual' has no property named 'original-background'

(ds-tao-segmentation:1278): GLib-GObject-WARNING **: 20:23:28.879: g_object_set_is_valid_property: object class 'GstNvSegVisual' has no property named 'alpha'
batchSize 1...
Failed to load config file: Key file does not start with a group
** ERROR: <gst_nvinfer_parse_config_file:1319>: failed
Now playing: /home/downloads/pgie_tlt_config_citysemsegformer.txt

Using winsys: x11 
No protocol specified
Running...

**PERF:  FPS 0 (Avg)	
Mon Dec 25 20:23:33 2023
**PERF:  0.00(0.00)	
Mon Dec 25 20:23:38 2023
**PERF:  0.00(0.00)	
Mon Dec 25 20:23:43 2023
**PERF:  0.00(0.00)	

Here are some warnings which I guess doesnt matter much (or they cannot impact since model cannot be loaded), but the main thing is that config file cannot be loaded for some reason. I used ones which are provided in guide, they did not work. In other deployments saw that jsons might also work - tried, didnt work.

Can you please help me to figure our how to make things working?

Please make sure this file is available and then set an absolute path and retry.

I am afraid it should be configs/nvinfer/citysemsegformer_tao/pgie_citysemsegformer_tao_config.txt now.

Refer to deepstream_tao_apps/configs/nvinfer/citysemsegformer_tao/pgie_citysemsegformer_tao_config.txt at master · NVIDIA-AI-IOT/deepstream_tao_apps · GitHub.

1 Like

Hello @Morgah! Thanks a lot for helping - the error was about config file. I was using correct paths, I suppose it was about this first line [property], I just havent seen this file before, it would be handy if guide would be referring to this file.

Also, I would like to mention for further readers - if you want to test model performance with -i option and input file -type file:// before input file - this is how to make things working.

1 Like

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.