custom-feature-extractor

A simple python script to extract features using deep learning models that are made available alongside pre-trained weights.

Available models

• ResNet50
• VGG19
• VGG16
• InceptionV3
• EfficientNetB0
• NASNetLarge

Instructions (How to use?)

1. Make sure that your image folder structure is organized as follows:

🔽 📁 my_images_dataset

▶️ 📁 images_1
▶️ ...
▶️ 📁 images_n

2. Call the python script scr_custom_feature_extractor.py followed by the required arguments:

• Arguments:
  • --directory - Data input directory
  • --model - One of available models: resnet50, vgg16, vgg19, inception_v3, efficient_net_b0, nas_large
  • --output_dir - Directory to save the output file
  • --file_type - Type of output file: txt, pkl or pbz2

An example:

python3 scr_custom_feature_extractor.py --directory "C:\my_images_dataset" --model vgg19 --output_dir "C:\my_features" --file_type pbz2 

⚠️ Work in progress...