Simple CNN is a pipeline which can be use to train and infer CNN models by use of PyTorch and ONNX. It's simple and easy to USE !!! 🔥🔥
- Clone the repo and install requirements.txt in a Python environment
git clone https://github.com/LahiRumesh/simple_cnn.git
cd simple_cnn
pip install -r requirements.txt- Split images into train and val folders with each class the Image Folder 📂.. i.e for cat vs dogs classification, there should be a cat folder and dog folder in both train and val. The following folder structure illustrates 3 classes
├── Image_Folder
├── train
│ │───── class1
│ │ ├── class1.0.jpg
│ │ ├── class1.1.jpg
│ │ ├── class1.2.jpg
│ │ ├── .........
│ │ └── class1.500.jpg
│ │
│ │───── class2
│ │ ├── class2.0.jpg
│ │ ├── class2.1.jpg
│ │ ├── class2.2.jpg
│ │ ├── .........
│ │ └── class2.500.jpg
│ │
│ └───── class3
│ ├── class3.0.jpg
│ ├── class3.1.jpg
│ ├── class3.2.jpg
│ ├── .........
│ └── class3.500.jpg
│
└── val
│───── class1
│ ├── class1.501.jpg
│ ├── class1.502.jpg
│ ├── class1.503.jpg
│ ├── .........
│ └── class1.600.jpg
│
│───── class2
│ ├── class2.501.jpg
│ ├── class2.502.jpg
│ ├── class2.503.jpg
│ ├── .........
│ └── class2.600.jpg
│
└───── class3
├── class3.501.jpg
├── class3.502.jpg
├── class3.503.jpg
├── .........
└── class3.600.jpg
- Use the config.py to set the parameters, here are few parameters.
cfg.data_dir = 'Data/Images/Image_Folder' # Image Folder path which contain train and val folders
cfg.device = '0' # cuda device, i.e. 0 or 0,1,2,3
cfg.image_size = 224 #input image size
cfg.batch_size = 8 # batch size
cfg.epochs = 50 #number of epochs
cfg.model = 'resnet18' # torch vision classification model architectures for image classification
# i.e. resnet18 or vgg16, alexnet, densenet121, squeezenet1_0
cfg.pretrained = True # use pretrained weights for training
#Early Stopping
cfg.use_early_stopping = True # use Early stopping
cfg.patience = 8 # how many epochs to wait before stopping when accuracy is not improving
cfg.min_delta = 0.0001 # minimum difference between new accuracy and old accuracy for new accuracy to be considered as an improvement -
Here are the Available pre-trained models in
Simple CNNArchitectures Available Models Resnet resnet18, resnet34, resnet50, resnet101, resnet152 VGG vgg13, vgg13_bn, vgg16, vgg16_bn, vgg19, vgg19_bn Densenet densenet121, densenet169, densenet161 , densenet201 Squeezenet squeezenet1_0, squeezenet1_1 Alexnet alexnet
Run cnn_train.py to start the training, all the logs will save in wandb, and ONNX weight files will save in the "models/Image_Folder" folder for each training experiment with the model name.
- After the training process, use the exported ONNX model for inference using cnn_inference.py
python cnn_inference.py --model_path=models/ImageFolder/ImageFolder_resnet18_exp_1.onnx --class_path=models/ImageFolder/classes.txt --img_path=test1.jpg --image_size=224 --use_transform=True'''
Args:
'''
--model_path : ONNX model path
--class_path : Class file (classes.txt) path contain class names
--img_path : Input image path
--image_size : input image size
--show_image : Display the image
--use_transform : Use image transforms in pre-processing step (During the training, process images are Normalize with a mean and standard deviation) - Use the test_accuracy.py to calculate the ONNX model accuracy on the test data.
python test_accuracy.py --model_path=models/ImageFolder/ImageFolder_resnet18_exp_1.onnx --class_path=models/ImageFolder/classes.txt --img_dir=Image_Folder/test --image_size=224 --use_transform=TrueThe following illustrates 3 classes of the test image folder
├── Image_Folder
├── test
│───── class1
│ ├── class1.0.jpg
│ ├── class1.1.jpg
│ ├── class1.2.jpg
│ ├── .........
│ └── class1.500.jpg
│
│───── class2
│ ├── class2.0.jpg
│ ├── class2.1.jpg
│ ├── class2.2.jpg
│ ├── .........
│ └── class2.500.jpg
│
└───── class3
├── class3.0.jpg
├── class3.1.jpg
├── class3.2.jpg
├── .........
└── class3.500.jpg
All the test results will save in the folder "test_results" folder for each test experiment.
'''
Args:
'''
--model_path : ONNX model path
--class_path : Class file (classes.txt) path contain class names
--img_dir : Test images folder path
--image_size : input image size
--use_transform : Use image transforms in pre-processing step (During the training, process images are Normalize with a mean and standard deviation) 
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent