Skip to content

install_macos-arm.md

Latest commit

 

History

History
57 lines (38 loc) · 1.86 KB

install_macos-arm.md

File metadata and controls

57 lines (38 loc) · 1.86 KB

FruitFinder v3

An AI fruit scanner for supermarket checkouts

  • Should probs do a complete re-think of middle model / training code using the eager_few_shot_od_training_tf2_colab.ipynb code as a base
  • Write model builder (/ loading from file) code (model.py)
  • Writing training loop

Usage Guide (MACOS M1 ARM)

Installation (CURRENTLY BROKEN)

  1. Install Anaconda
  2. Create a clean conda environment and activate it
  3. Install all of the required packages using conda env create -f environment.yml --name fruitfinder (see full dependancy list below)
  4. Download the resnet50 model from here. Note: various other COCO models accessible here

Installing the Object Detection API

cd models/research/
protoc object_detection/protos/*.proto --python_out=.
cp object_detection/packages/tf2/setup.py .
python -m pip install --no-deps  .

Usage

  • Activate the conda environment by running conda activate fruitfinder
  • Run python train.py to train the model
  • Run python predict.py to test out the trained model

Dependancies

The following main dependancies were used in the project (see environment.yml to see full dependancy list):

  • tensorflow (version 2.9.2)
  • tensorflow_probability (version 0.17.0)
  • numpy (version 1.23.3)
  • matplotlib (version 3.5.1)
  • PIL / pillow (version 9.1.0)
  • imageio (version 2.22.1)
  • skimage (version 0.19.3)
Made with ❤️