How to run object_detection_demo_faster_rcnn using tensorfllow's faster rcnn resnet101 model

[Taiki]

I want to run object_detection_demo_faster_rcnn using an IR file converted from a tensorfllow faster rcnn resnet101 model, but I don't know what to specify for the "bbox_pred", "proposal" and "ccls_prob" options.

[Rizal_Intel]

Hi Taiki,

You would need to search inside your model using Netron application to find corresponding layer names for the arguments.

For example, the pre trained model faster-rcnn-resnet101-coco-sparse-60-0001.

The layer names are as follows:

proposal_name = proposals

bbox_name = div

prob_name = openvino_outputs/cls_score

You can try searching for a particular layer name using Netron.

In the example the next layer contains name such as bbox_reshape which helps to identify the previous layer "div" as the bbox result layer.

Regards,

Rizal

[Munesh_Intel]

Hi Taiki,

This thread will no longer be monitored since we have provided steps to solve your issue. If you need any additional information from Intel, please submit a new question.

Regards,

Munesh