Views
No views yet
pip install -r requirements.txtfas-datasets/
├── publics_data_test/
│ └── data_test/
│ ├── 10.mp4
│ ├── 108.MOV
│ ├── 11.mp4
│ └── ... (video files)
└── publics_data_train/
├── Galaxy_A53/
├── user007/
├── user020/
├── user034/
├── user035/
├── users_imgs/
├── spoof/
└── live/data/processed/
├── train/
│ ├── Galaxy_A53/
│ ├── user007/
│ ├── video_frames/
│ └── ...
├── val/
└── test/1# Automatic processing of your dataset structure
2python efficientnet_classifier.py
3
4# The script will automatically:
5# 1. Extract frames from videos in data_test
6# 2. Process image folders from publics_data_train
7# 3. Combine and clean the data
8# 4. Split into train/val/test
9# 5. Train the model
10
11# For prediction with class names:
12import json
13with open('class_mapping.json', 'r') as f:
14 class_mapping = json.load(f)
15
16from efficientnet_classifier import EfficientNetClassifier, predict_single_image
17model = EfficientNetClassifier(num_classes=len(class_mapping['classes']))
18model.load_state_dict(torch.load('best_efficientnet_b3.pth'))
19predict_single_image(model, "test_image.jpg", class_mapping['idx_to_class'])BASE_DATA_DIR: Path to your fas-datasets folderFRAMES_PER_VIDEO: Number of frames to extract per video (default: 10)VideoFrameExtractor