Ticket Name: TDA2SX: PROCESSOR_SDK_VISION_03_05 tidl_OD usecase layer limitations ?

Query Text:
Part Number: TDA2SX Hi I tried to use tidl_model_import.out to get bin file. I used stats_tool_out.bin and trace_dump_0_512x512.y file to check test image detection result. Test image showed model can detect object. test image detection result: But I ran on tdasx, it cannot detect any object. I suspect tdasx cannot run too many layers model. The number of layersGroupId are 157 and number of parameters are 3.5M. Do tdasx have any limitations about layers or parameters count? thanks yumei

Responses:
Hi Yumei, Below are the SSD limitations listed in the TIDL datasheet : – Only Caffe-Jacinto based SSD network is validated. – Reshape, Permute layers are supported only in the context of SSD network. – “share_location” has to be true – Tested with 4 and 5 heads. – SaveOutputParameter is ignored in TIDL inference. – code_type is only tested with CENTER_SIZE. Please try the suggestions mentioned in the below thread in OD use case to get the detections in the output : https://e2e.ti.com/support/processors-group/processors/f/processors-forum/689617/tda2-how-to-run-ssd-based-tidl-od-use-case-in-vision-sdk-with-pre-trained-model Thanks, Praveen

Hi Praveen I confused about Tested with 4 and 5 heads. Did it mean about 4 and 5 anchor box result? Thanks, yumei

Yes