muka / tensorflow-experiments

Tensorflow based, good apples image recognition

Apple type recognition

Notes for experiments on tensorflow and apple recognition

Setup

• Get models make fetch-models
• Ensure tensroflow py modules are reachable export PYTHONPATH=$PYTHONPATH:${PWD}/models/research:${PWD}/models/research/slim
• Run TF make run-tensorflow

Image labelled with MS VoTT https://github.com/Microsoft/VoTT.git

Testing

I used that repo https://github.com/chtorr/go-tensorflow-object-detection to get RT detection out of Webcam

• Clone & install as descripted in the repo
• Edit under data/ setting 1: to apple
• launch specifying the directory where you exported the frozen model (see make export-trainer)

References

• https://gist.github.com/douglasrizzo/c70e186678f126f1b9005ca83d8bd2ce
• https://medium.com/@WuStangDan/step-by-step-tensorflow-object-detection-api-tutorial-part-4-training-the-model-68a9e5d5a333
• https://towardsdatascience.com/how-to-train-your-own-object-detector-with-tensorflows-object-detector-api-bec72ecfe1d9
• https://github.com/tensorflow/models/blob/master/research/object_detection/g3doc/using_your_own_dataset.md