/EasyAI

Make your own AI easily !

Primary LanguagePythonMIT LicenseMIT

EasyAI

Make your own AI easily ! via tensorflow 2

EasyAI

Description

The main goal of this project is to create an AI easily using tensorflow

Software requirements

Hardware requirements

We recommend using a GPU with Hardware Acceleration for Tensorflow.

Hardware minimum recommended
CPU 2 cores 4 cores
RAM 4 GB 8 GB
GPU 2 GB 4 GB

Usage

Get help:

python3 EasyAI.py --help

Launch training:

python3 EasyAI.py

Launch training without GPU:

python3 EasyAI.py --no-gpu

Launch training, save model and set dataset path:

python EasyAI.py --data_dir ./dataset/flower_photos --save test_AI.h5

Example

Train a model with flowers dataset:

python EasyAI.py --save test_AI.h5

Predict images (class_name order is important):

python EasyAI.py --load test_AI.h5 --predict ./dataset/flower_photos/roses/ --class_name daisy dandelion roses sunflowers tulips

Command table

Command Description Default Example
--data_dir Path to the data directory. ./dataset/ --data_dir ./dataset/
--save-model Save model to a HDF5 file. None --save test_AI.h5
--load-model Load model from a HDF5 file. None --load test_AI.h5
--predict Predict images. None --predict ./dataset/flower_photos/roses/
--class_name Class name. None --class_name daisy dandelion roses sunflowers tulips
--no-gpu Disable GPU. False --no-gpu
--batch_size Batch size. 32 --batch_size 32
--epochs Number of epochs. 10 --epochs 10
--model_path Path to the model. None --model_path ./model/
--tensorboard Enable tensorboard (Slow). False --tensorboard
--checkpoint Enable checkpoint. False --checkpoint

Done features

  • Working model
  • Tensorboard integration
  • GPU support
  • Load and save model
  • Data augmentation
  • Argument parser

Work in progress features

  • Data set generator
  • Load and save weights

Future features

  • Lite version
  • Docker image
  • pip package

Open source projects used

Licensing

MIT License