Unable to load may MyModel.dnn in C#

[aminullah6264]

i have train my object detection model on FastRCNN using python and save it using save_model("MyModel.dnn").
Now i try to wright script to evaluate it in C#. it throw an exception on following encircle line.
that Error: Sequence contains more than one element.
please help me to solve this problem problem.
This is my code for loading model, i have follow the steps, which is given in CNTK examples

[pkranen]

Fast R-CNN requires two inputs (hence "Sequence contains more than one element"), the first is the image and the second are the ROI proposals (see this tutorial for more details). For evaluation in Python see evaluate_fast_rcnn() method in A2_RunWithPyModel.py:

    input_map = {
        model.arguments[0]: test_minibatch_source[features_stream_name],
        model.arguments[1]: test_minibatch_source[roi_stream_name],
    }

In C# you would need to create the ROI proposals yourself. We do not have a C# implementation.

[aminullah6264]

Dear @pkranen
i don't have ROIs for Test images.
i train my model on images of 10 videos
now i want to test it for more videos in which i don't have ROIs
i want that model should predict ROIs for me

[pkranen]

The model does not predict ROIs, it requires ROIs as an input and predicts whether and which object is in the ROI. You can try Faster R-CNN which has a region proposal network and does not require ROIs as an input. There will be a tutorial added on Faster R-CNN in the coming weeks, for now please use the Readme file in the above folder.

[aminullah6264]

Dear @pkranen
can i train FasterRCNN to detect ROIs using the following setups given in your Examples.