Can deepsteram5.0 encode the selected frames into files?

Please provide complete information as applicable to your setup.

• Hardware Platform (Jetson / GPU) Tesla T4
• DeepStream Version 5.0
• JetPack Version (valid for Jetson only) No
• TensorRT Version save as deepstream:5.0-dp-20.04-devel
• NVIDIA GPU Driver Version (valid for GPU only) 10.2

I want to encode the decoded video frames in the buffer into jpeg format, and save it into files. I found that “deepstream_image_meta_test” sample application supports save corped objects into image files. Can Nvidia provide some api to encode the whole frame or open source the “nvds_obj_enc_process” implementation so that I can do it myself?

The nvds_obj_enc_process API can also be used to encode the whole frame, just set your own object meta as the whole picture then the whole picture will be encoded to picture.

https://docs.nvidia.com/metropolis/deepstream/dev-guide/DeepStream_Development_Guide/baggage/group__ee__object__encoder.html

Could you please offer me a simple demo of this? I am stucked here…

Do I need to acquire an object_meta from pool, set it’s class id, rect_params, and then add it to frame_meta before calling "nvds_obj_enc_process " API?

What is the purpose of “save_img” and “attach_user_meta” in NvDsObjEncUsrArgs? What will the nvds_obj_enc_process API behave if I set them to FALSE?

There is no sample codes, but it is OK to acquire an object_meta from pool and set only one object for the whole image with the new object meta.

“save_img” is to enable save the object as image file, “attach_user_meta” is to enable attaching the image to the user meta and attaching the user meta to the object meta. https://docs.nvidia.com/metropolis/deepstream/dev-guide/DeepStream_Development_Guide/baggage/struct__NvDsObjEncUsrArgs.html

1 Like