image-model

ML image model for detecting vehicles in an image

Setup development environment (BRANCH PLEASE)

---

Clone the repo using:

git clone https://github.com/EggHeads-cs35l/image-model.git

For the rest of the setup, navigate to the cloned directory and run commands there.

We will be using Conda (Miniconda) to create our development environment: https://anaconda.org

Next, update conda by running:

conda update -n base -c defaults conda

Step 1: Install Miniconda

Navigate to the miniconda installer page and download your correpsonding file: https://docs.conda.io/en/latest/miniconda.html

The process will enitrely depend on your development platform: Windows, WSL, MacOS, Linux.

Regardless of platform, we recommend installing via the bash script: simply run the downloaded script with the prefix bash <FILE>

Step 2: Create Conda environment

Refer to the Conda cheat sheet for help with commands: https://docs.conda.io/projects/conda/en/4.6.0/_downloads/52a95608c49671267e40c689e0bc00ca/conda-cheatsheet.pdf

Run the following commands to setup a conda environment:

Make sure to replace "<NAME>" with the actual name of the environemnt you wish to create.

conda create --name <NAME> python=3.9

We will be using Python 3.9.

You can activate and deactivate your conda environment using:

conda activate <NAME>
conda deactivate

or

source activate <NAME>
source deactivate

---

IMPORTANT: From now on, make sure you activate your conda envrironment

Step 3 (Only for MacOS): Set Up Apple Metal

If you are using MacOS, we recommend using Apple Metal to enhance training using your GPU on Intel Macs or GPU cores on the new Apple Silicon: https://developer.apple.com/metal/tensorflow-plugin/

If you are on an Apple silicon Mac, run the following command first:

conda install -c apple tensorflow-deps

If you are on MacOS (either Intel or Apple silicon), run these commands:

install TensorFlow for MacOS

python -m pip install tensorflow-macos

install Metal fro TensorFlow

python -m pip install tensorflow-metal

Step 4: Install dependencies

We will be using TensorFlow for our ML framework: https://www.tensorflow.org

Run the following command (in the directory) sequentially to install dependencies:

conda install numpy pandas matplotlib
conda install -c conda-forge tensorflow pillow
conda install -c anaconda ipykernel

Step 5: Donwload the dataset

We will use the Stanford Cars Dataset: http://ai.stanford.edu/~jkrause/cars/car_dataset.html

3D Object Representations for Fine-Grained Categorization

Jonathan Krause, Michael Stark, Jia Deng, Li Fei-Fei

4th IEEE Workshop on 3D Representation and Recognition, at ICCV 2013 (3dRR-13). Sydney, Australia. Dec. 8, r013.

run all of the following commands sequentially:

mkdir data
cd data
curl http://ai.stanford.edu/~jkrause/car196/car_ims.tgz > images.tgz
curl http://ai.stanford.edu/~jkrause/car196/cars_annos.mat > info.mat
tar -xzf car_ims.tgz -C images/

Step 6: Organize data

We will create directories used later now so we don't accidentally create redundant directories run all of the following commands in the terminal (in this repo's directory):

cd data
mkdir dataset
cd dataset
mkdir {0..195}

Step 7: Sanitize your commits

To prevent from committing outputs of Jupyter code blocks run the following command:

git config filter.strip-notebook-output.clean 'jupyter nbconvert --ClearOutputPreprocessor.enabled=True --to=notebook --stdin --stdout --log-level=ERROR'

Now you are ready to develop!