Skip to content

process of plugging in a custom Tensorflow model #19

Open
@RuchiSaluja8

Description

@RuchiSaluja8

Hi
I am experimenting with the Rocket Platform and ran pipeline 4, but got the following output :

~$ sudo docker run --runtime=nvidia -v /home/Createlabs/Microsoft-Rocket-Video-Analytics-Platform/output:/app/output lvareg.azurecr.io/custom_rocket sample.mp4 sample.txt 1 1 tree
./cfg/sample.txt
Downloading default label map
Downloading default model
Extracting
dir = /app

Unhandled Exception: System.IO.DirectoryNotFoundException: Could not find a part of the path '/app/faster_rcnn_openimages_v4_inception_resnet_v2_1/frozen_inference_graph.pb'.
   at Interop.ThrowExceptionForIoErrno(ErrorInfo errorInfo, String path, Boolean isDirectory, Func`2 errorRewriter)
   at Microsoft.Win32.SafeHandles.SafeFileHandle.Open(String path, OpenFlags flags, Int32 mode)
   at System.IO.FileStream..ctor(String path, FileMode mode, FileAccess access, FileShare share, Int32 bufferSize, FileOptions options)
   at System.IO.FileStream..ctor(String path, FileMode mode, FileAccess access, FileShare share, Int32 bufferSize)
   at System.IO.File.ReadAllBytes(String path)
   at Wrapper.TF.TFWrapper..ctor() in /app/src/VAP/TFWrapper/TFWrapper.cs:line 64
   at TFDetector.FrameDNNTF..ctor(List`1 lines) in /app/src/VAP/TFDetector/FrameDNNTF.cs:line 25
   at TFDetector.LineTriggeredDNNTF..ctor(List`1 lines) in /app/src/VAP/TFDetector/LineTriggeredDNNTF.cs:line 25
   at VideoPipelineCore.Program.Main(String[] args) in /app/src/VAP/VideoPipelineCore/Program.cs:line 75

(The sample.mp4 is a video containing trees.)

Currently, I plugged in a model from Tensorflow Hub and have changed lines 116-119 in src/VAP/TFWrapper/TFWrapper.cs to :

string defaultModelUrl = "https://tfhub.dev/google/faster_rcnn/openimages_v4/inception_resnet_v2/1?tf-hub-format=compressed";

            var modelFile = Path.Combine(dir, "faster_rcnn_openimages_v4_inception_resnet_v2_1/frozen_inference_graph.pb");
            var zipfile = Path.Combine(dir, "faster_rcnn_openimages_v4_inception_resnet_v2_1.tar.gz");

The error indicates looking for the frozen_inference_graph.pb, which is a file in the original model .

Please share the process of plugging in a custom Tensorflow model.
I am new to ML and the Rocket platform. Any help would be appreciated. TIA!

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions