search

experiencor / keras-yolo2

Easy training on custom dataset. Various backends (MobileNet and SqueezeNet) supported. A YOLO demo to detect raccoon run entirely in brower is accessible at https://git.io/vF7vI (not on Windows).
MIT License
1.73k stars 784 forks source link

How to detect on an image using C# language #381

Open smartyfn opened 6 years ago

smartyfn commented 6 years ago

Thanks for your good source code. I am a C# engineer. I don't want to use the predict.py file . I know tensorflowsharp but it doesn't have C# api of keras. How to detect on an image using C# language. What is the best way to solve it? Any suggestions will be greatly appreciated.

rodrigo2019 commented 6 years ago

convert the .hdf5 model from keras to .pb from tensorflow and run the graph in your C# api. But you will need implement the decode_netout function in c#

here is a example how to convert the models: https://github.com/amir-abdi/keras_to_tensorflow

smartyfn commented 6 years ago

@rodrigo2019 Thank you very much,I will try it