In deepstream-5.0, need to save analysed image frame if any object gets detected in a frame. I was referring “Apps/sample_apps / deepstream-image-meta-test” as was suggested in few threads but it only emits cropped image which has the object.
Also tried to intercept the meta being transferred from upstream as _NvDsMetaPool, but not able to understand how to get desired image info. from it. It contains “empty_list” and “full_list” which I fail to make sense of. And in any of those list the present _NvDsObjectMeta object always give -0.1 confidence. I believe that struct do have all info. which I might need to construct a full frame, boundary boxed image, but not able to understand it.
Would be great if someone can put me on right track. It’s important for our use-case to save full-frame image snapshot upon any detection.