ssd_inception_v2_coco custom trained model with my own dataset. about tensorrt_demos

[jkjung-avt]

You could reference NVIDIA's original code: model_ssd_inception_v2_coco_2017_11_17.py

The modification should go into build_engine.py in this repository.

[SahilChachere]

Solved it by replacing input: "image_tensor: 0" by input: "Cast".

graph.collapse_namespaces(namespace_plugin_map)
    graph.remove(graph.graph_outputs, remove_exclusive_dependencies=False)
    graph.find_nodes_by_op("NMS_TRT")[0].input.remove("Input")
    #if model == 'ssd_inception_v2_coco':
        #graph.find_nodes_by_name("Input")[0].input.remove("image_tensor:0")
        #graph.find_nodes_by_name("Input")[0].input.remove("Cast")

try above changes in 'build_engine.py' file @maky-hnou. I have commented 3 lines and it's working as it is not required. also can you put your error here?

[SahilChachere]

Solved it by replacing input: "image_tensor: 0" by input: "Cast".

[maky-hnou]

@SahilChachere could you please mention in which file did you make that change?