tiovxdlcolorblend: Not working in EdgeAI apps

[aniket-l]

PFA error logs when integrating the plugin into the EdgeAI apps.
blend_error_logs.txt

Also PFA a patch file
colorblend.txt
to be applied on top of master branch commit id 969a319 (https://git.ti.com/cgit/edgeai/edge_ai_apps/commit/?id=969a319acb8a4b112f4e3176fdc96abc9f16a6fb) to reproduce the error.
Also note that in the config file semantic_segmentation.yaml, use "models: [model1]" in the flow as the other models use int64 data type which is not enabled yet.

[jafet-chaves]

Aniket, as mentioned previously the solution to this issue is that (at least in the commit 969a319acb8a4b112f4e3176fdc96abc9f16a6fb) the following pipeline branch needs to specify the data type, width and height. This allows caps negotiation to work. Please check the diff below

diff --git a/apps_python/gst_wrapper.py b/apps_python/gst_wrapper.py
index 0494488..13603e0 100644
--- a/apps_python/gst_wrapper.py
+++ b/apps_python/gst_wrapper.py
@@ -497,7 +497,7 @@ def get_post_proc_str(flow):
         sen_cmd += '! blend.sink \t'
         infer_cmd = 'appsrc format=GST_FORMAT_TIME is-live=true block=true ' + \
                     'do-timestamp=true name=%s ' % flow.gst_infer_sink_name + \
-                    '! application/x-tensor-tiovx ' + \
+                    '! capssetter ! application/x-tensor-tiovx,data-type=6,num-dims=3,tensor-width=512,tensor-height=512 ' + \
                     '! queue ! blend.tensor \t'
         sink_cmd = 'tiovxdlcolorblend name=blend target=1 num-classes=%d data-type=%d ' % (flow.model.num_classes,data_type) + \
                     '! video/x-raw,format=NV12 '+ \