Three cameras on Jetson Nano

[DaneLLL]

Hi,
Performance of deep learning inference depends on the models. We have a sample config file for Jetson Nano:

/opt/nvidia/deepstream/deepstream-5.0/samples/configs/deepstream-app/source8_1080p_dec_infer-resnet_tracker_tiled_display_fp16_nano.txt

The model is resnet10.

For sources in MJPEG, you may refer to this patch:

How to change source pixel format from YUYV to MJPG

Hi DaneLLL, Finally, it works. I really appreciate your help. bin->cap_filter = gst_element_factory_make (NVDS_ELEM_CAPS_FILTER, "src_cap_filter"); if (!bin->cap_filter) { NVGSTDS_ERR_MSG_V ("Could not create 'src_cap_filter'"); goto done; } caps = gst_caps_new_simple ("image/jpeg", "width", G_TYPE_INT, config->source_width, "height", G_TYPE_INT, config->source_height, "framerate", GST_TYPE_FRACTION, config->source_fps_n, config->source_fps_d, NULL); …

It uses jpegdec in the patch. Another working solution is to use nvv4l2decoder mjpeg=1.