Image taken via tiler_sink_pad_probe is damaged

Please provide complete information as applicable to your setup.

• Hardware Platform (Jetson / GPU) Jetson Nano
• DeepStream Version 6.0.1
• JetPack Version (valid for Jetson only) 4.6.1-b110
• TensorRT Version 8.2

I use similar code as shown in in deepstream_python_apps/apps/deepstream-imagedata-multistream/ at v1.1.1 · NVIDIA-AI-IOT/deepstream_python_apps · GitHub

The simplified code of tiler_sink_probe is

gst_buffer = info.get_buffer()

buffer_hash = hash(gst_buffer)
batch_meta = pyds.gst_buffer_get_nvds_batch_meta(buffer_hash)

l_frame = batch_meta.frame_meta_list

while l_frame is not None:
frame_meta = pyds.NvDsFrameMeta.cast(
except StopIteration:

cv_frame_buf = pyds.get_nvds_buf_surface(buffer_hash, frame_meta.batch_id)
cv_frame = np.array(cv_frame_buf, copy=True, order='C')
cv_frame = cv2.cvtColor(cv_frame, cv2.COLOR_RGBA2BGR) #A)

cv2.imwrite(f"/tmp/img_{frame_meta.frame_num}.jpg", cv_frame)

but the saved images are still broken. Please, could you help me somehow?

Are you using a rtsp source or a local file source? From the image attached, it looks like a network packet loss in rtsp.

No, I use the CSI camera as a source. And, a few of the first images look like the following:

OK. Could you save the output to video by referring to our FAQ? We can rule out whether opencv is the cause of the issue by that.

The output shown on my monitor is completely OK. Problematic are only pictures saved via the tiler_sink_probe. Also, another strange thing is that on my laptop the code works correctly and pictures saved via the tiler_sink_probe are OK too. Only Jetson Nano seems to be somehow problematic

What exactly do you mean by the image damage? Can you describe it in detail and attach the original image?

There is no update from you for a period, assuming this is not an issue anymore. Hence we are closing this topic. If need further support, please open a new one. Thanks

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