Error trying to run object_detection_sample_ssd

[js07920]

The error is "Can't find a DetectionOutput layer in the topology"

Using OV 2020.2

Download and conversion steps taken:

python \intel\openvino\openvino\deployment_tools\tools\model_downloader\downloader.py --name alexnet -o C:\Users\user\Documents\Intel\OpenVINO\openvino_models\

python \intel\openvino\openvino\deployment_tools\model_optimizer\mo.py --input_proto C:\Users\user\Documents\Intel\OpenVINO\openvino_models\public\alexnet\alexnet.prototxt  --input_model C:\Users\user\Documents\Intel\OpenVINO\openvino_models\public\alexnet\alexnet.caffemodel

python object_detection_sample_ssd.py -m alexnet.xml -i n:\intel\001_1024.jpeg
[ INFO ] Loading Inference Engine
[ INFO ] Loading network files:
alexnet.xml
alexnet.bin
[ INFO ] Device info:
CPU
MKLDNNPlugin version ......... 2.1
Build ........... 42025
[ INFO ] File was added:
[ INFO ] n:\intel\001_1024.jpeg
[ WARNING ] Image n:\intel\001_1024.jpeg is resized from (227, 227) to (227, 227)
[ INFO ] Preparing input blobs
[ INFO ] Batch size is 1
[ INFO ] Preparing output blobs
[ ERROR ] Can't find a DetectionOutput layer in the topology
[ ERROR ] Incorrect output dimensions for SSD model
Traceback (most recent call last):
File "object_detection_sample_ssd.py", line 188, in <module>
sys.exit(main() or 0)
File "object_detection_sample_ssd.py", line 132, in main
max_proposal_count, object_size = output_dims[2], output_dims[3]
IndexError: list index out of range

C:\Users\user\Documents\Intel\OpenVINO\jay_changes\object_detection_sample_ssd

[Vladimir_Dudnik]

Hello,

I would recommend you to use 

\intel\openvino\openvino\deployment_tools\tools\model_downloader\converter.py --name alexnet -o C:\Users\user\Documents\Intel\OpenVINO\openvino_models\

in addition to downloader.py script. This converter.py script use MO conversion parameters, specified for each OMZ public model in model.yml file, and conversion to IR is validated each OpenVINO release. So this is method quarantined to work, while trying to convert model manually can lead to different mistakes and as a result to failure during conversion.

Regards,
  Vladimir

[SIRIGIRI_V_Intel]

Hi Js,

Alexnet is a classification model and is not a suggested model for object_detection_ssd demo. Please use object detection model(person-detection-retail-0013, mobilenet-ssd, etc.) for this demo.

Regards,

Ram prasad