search

javathunderman / diabetic-retinopathy-screening

Diabetic retinopathy screening w/ Tensorflow.
GNU General Public License v3.0
68 stars 33 forks source link

tf2onnx - how? #4

Closed AndreV84 closed 2 years ago

AndreV84 commented 4 years ago

could you help out figuring flags for converting the retrained.pb to onnx, please? https://forums.developer.nvidia.com/t/implementing-deepstream-trt-integration-by-intels-scenario/153553

javathunderman commented 4 years ago

Hi there,

This project is quite old at this point - much of the code hasn't been touched in about 3 years and there have been significant revisions made to the underlying APIs since then. At some point, I will probably update this project and add new features to it, but in the meantime, I highly suggest following Google's image classifier tutorial. Training MobileNet/ImageNet on diabetic retinopathy images should be fairly straightfoward if you follow that tutorial (basically swap the dataset they provide for the DR dataset).

Sorry that I can't be more helpful. Please let me know if you have other questions - I'll try to answer as much as I can. Thanks.

AndreV84 commented 4 years ago

Thank you for following up! I shall try "MobileNet/ImageNet on diabetic retinopathy images" I will post question sa thy emerge if you do not mind.

AndreV84 commented 4 years ago

@javathunderman , If I setup access to a cloud instance with dGPU, could you try to train a model on it, please? So that it will be an updated model? So that it could be possible further to approach integrating it into TensorRT with further integration into DeepStream? Reference threads https://forums.developer.nvidia.com/t/implementing-deepstream-trt-integration-by-intels-scenario/153553 https://forums.developer.nvidia.com/t/given-there-is-engine-file-h5-how-to-incorporate-it-into-deepstream/145401/35