Fix dependency management

.gitignore

@@ -5,3 +5,4 @@
 *.pth
 *.pt
 *.dat
+.envrc

preprocessing/config.yaml

@@ -24,10 +24,10 @@ rearrange:
     default: 'FaceForensics++'
   dataset_root_path: # the root path to the dataset
     type: str
-    default: ''
+    default: 'datasets/rgb'
   output_file_path: # the json path to the dataset
     type: str
-    default: '../preprocessing/dataset_json_v6'
+    default: 'datasets/json'
   comp: # the compression level of videos, only in the dataset of FaceForensics++.
     choices: ['raw', 'c23', 'c40']
     default: 'c23'

preprocessing/rearrange.py

@@ -500,7 +500,7 @@ def generate_dataset_file(dataset_name, dataset_root_path, output_file_path, com
 
 if __name__ == '__main__':
     # from config.yaml load parameters
-    yaml_path = './config.yaml'
+    yaml_path = './preprocessing/config.yaml'
     # open the yaml file
     try:
         with open(yaml_path, 'r') as f:

pyproject.toml

@@ -0,0 +1,44 @@
+[project]
+name = "deepfakebench"
+version = "0.1.0"
+description = "Add your description here"
+readme = "README.md"
+requires-python = ">=3.10"
+dependencies = [
+    "albumentations>=2.0.8",
+    "clip",
+    "dlib>=20.0.0",
+    "efficientnet-pytorch>=0.7.1",
+    "einops>=0.8.1",
+    "filterpy>=1.4.5",
+    "fvcore>=0.1.5.post20221221",
+    "imageio>=2.37.0",
+    "imgaug>=0.4.0",
+    "imutils>=0.5.4",
+    "kornia>=0.8.1",
+    "loralib>=0.1.2",
+    "numpy<2.0",
+    "opencv-python>=4.11.0.86",
+    "pandas>=2.3.0",
+    "pillow>=11.2.1",
+    "pyyaml>=6.0.2",
+    "requests>=2.25.0",
+    "scikit-image>=0.25.2",
+    "scikit-learn>=1.7.0",
+    "scipy>=1.15.3",
+    "seaborn>=0.13.2",
+    "segmentation-models-pytorch>=0.5.0",
+    "setuptools>=80.9.0",
+    "simplejson>=3.20.1",
+    "tensorboard>=2.19.0",
+    "timm>=1.0.16",
+    "torch>=2.7.1",
+    "torchaudio>=2.7.1",
+    "torchtoolbox>=0.1.8.2",
+    "torchvision>=0.22.1",
+    "tqdm>=4.67.1",
+    "transformers>=4.53.0",
+]
+
+[tool.uv.sources]
+clip = { git = "https://github.com/openai/CLIP.git" }

training/config/train_config.yaml

@@ -1,9 +1,9 @@
 mode: train
-lmdb: True
+lmdb: False
 dry_run: false
-rgb_dir: './datasets/rgb'
+rgb_dir: '.'
 lmdb_dir:  './datasets/lmdb'
-dataset_json_folder: './preprocessing/dataset_json'
+dataset_json_folder: './datasets/json'
 SWA: False
 save_avg: True
 log_dir: ./logs/training/

training/dataset/abstract_dataset.py

@@ -14,6 +14,7 @@
 import yaml
 import glob
 import json
+from pathlib import Path
 
 import numpy as np
 from copy import deepcopy
@@ -115,11 +116,11 @@ def init_data_aug_method(self):
             A.HorizontalFlip(p=self.config['data_aug']['flip_prob']),
             A.Rotate(limit=self.config['data_aug']['rotate_limit'], p=self.config['data_aug']['rotate_prob']),
             A.GaussianBlur(blur_limit=self.config['data_aug']['blur_limit'], p=self.config['data_aug']['blur_prob']),
-            A.OneOf([                
-                IsotropicResize(max_side=self.config['resolution'], interpolation_down=cv2.INTER_AREA, interpolation_up=cv2.INTER_CUBIC),
-                IsotropicResize(max_side=self.config['resolution'], interpolation_down=cv2.INTER_AREA, interpolation_up=cv2.INTER_LINEAR),
-                IsotropicResize(max_side=self.config['resolution'], interpolation_down=cv2.INTER_LINEAR, interpolation_up=cv2.INTER_LINEAR),
-            ], p = 0 if self.config['with_landmark'] else 1),
+            # A.OneOf([                
+            #     IsotropicResize(max_side=self.config['resolution'], interpolation_down=cv2.INTER_AREA, interpolation_up=cv2.INTER_CUBIC),
+            #     IsotropicResize(max_side=self.config['resolution'], interpolation_down=cv2.INTER_AREA, interpolation_up=cv2.INTER_LINEAR),
+            #     IsotropicResize(max_side=self.config['resolution'], interpolation_down=cv2.INTER_LINEAR, interpolation_up=cv2.INTER_LINEAR),
+            # ], p = 0 if self.config['with_landmark'] else 1),
             A.OneOf([
                 A.RandomBrightnessContrast(brightness_limit=self.config['data_aug']['brightness_limit'], contrast_limit=self.config['data_aug']['contrast_limit']),
                 A.FancyPCA(),
@@ -296,7 +297,7 @@ def load_rgb(self, file_path):
         size = self.config['resolution'] # if self.mode == "train" else self.config['resolution']
         if not self.lmdb:
             if not file_path[0] == '.':
-                file_path =  f'./{self.config["rgb_dir"]}\\'+file_path
+                file_path =  f'./{self.config["rgb_dir"]}/'+file_path
             assert os.path.exists(file_path), f"{file_path} does not exist"
             img = cv2.imread(file_path)
             if img is None:
@@ -469,14 +470,21 @@ def __getitem__(self, index, no_norm=False):
         label = self.data_dict['label'][index]
 
         if not isinstance(image_paths, list):
-            image_paths = [image_paths]  # for the image-level IO, only one frame is used
+            image_paths = [Path(image_paths)]  # for the image-level IO, only one frame is used
+
+        image_paths = [Path(path) for path in image_paths]  # Sanitize image paths
 
         image_tensors = []
         landmark_tensors = []
         mask_tensors = []
         augmentation_seed = None
 
         for image_path in image_paths:
+            # check if this is a valid path
+            if not image_path.exists():
+                raise ValueError(f"Image path {image_path.absolute()} does not exist.")
+
+            image_path = str(image_path)
             # Initialize a new seed for data augmentation at the start of each video
             if self.video_level and image_path == image_paths[0]:
                 augmentation_seed = random.randint(0, 2**32 - 1)

training/test.py

@@ -37,7 +37,7 @@
 
 parser = argparse.ArgumentParser(description='Process some paths.')
 parser.add_argument('--detector_path', type=str, 
-                    default='/home/zhiyuanyan/DeepfakeBench/training/config/detector/resnet34.yaml',
+                    default='config/detector/resnet34.yaml',
                     help='path to detector YAML file')
 parser.add_argument("--test_dataset", nargs="+")
 parser.add_argument('--weights_path', type=str, 

training/train.py

@@ -38,7 +38,7 @@
 
 parser = argparse.ArgumentParser(description='Process some paths.')
 parser.add_argument('--detector_path', type=str,
-                    default='/data/home/zhiyuanyan/DeepfakeBenchv2/training/config/detector/sbi.yaml',
+                    default='training/config/detector/xception.yaml',
                     help='path to detector YAML file')
 parser.add_argument("--train_dataset", nargs="+")
 parser.add_argument("--test_dataset", nargs="+")