google / CFU-Playground

Want a faster ML processor? Do it yourself! -- A framework for playing with custom opcodes to accelerate TensorFlow Lite for Microcontrollers (TFLM). . . . . . Online tutorial: https://google.github.io/CFU-Playground/ For reference docs, see the link below.
http://cfu-playground.rtfd.io/
Apache License 2.0
475 stars 120 forks source link

Trying to inference the image in Arty A7 35T board. #706

Open Manikandanbrema opened 2 years ago

Manikandanbrema commented 2 years ago

Hello, We are trying to run an model in CFU playground. We followed the steps we can able to run pdti8 model in ArtyA7 board but what we want to do is to send our custom model which we trained and converted into Tensorflow lite into the CFU playground and an image into it and want to inference the image with that model and accelerating the model ops. We are new to this part but interested to do it.

Could you please guide us in a right path to complete this.

tcal-x commented 2 years ago

Hello @Manikandanbrema. We don't have a tutorial for adding a model since it can vary depending on the model. But generally, you need to add the model itself (in .tflite form), one or more input datasets, correct output for each data set, and finally modify the CFU Playground code to add menu options to load the model and run the tests. That last step is not so difficult since it's mostly cut and paste and modify.

You can use earlier PRs such as this one #111 as a guide, and even look at the individual commits. I think in this one Joey added all of the models in one commit, then added the inputs/outputs/code for each model in individual following commits.