Merge branch 'f/fix_vision_deps' into 'main'

Fix dependency on computer vision datasets

See merge request es/ai/hannah/hannah!353

hannah/conf/module/anomaly_detection.yaml

@@ -1,4 +1,4 @@
-_target_: hannah.modules.AnomalyDetectionModule
+_target_: hannah.modules.vision.AnomalyDetectionModule
 num_workers: 0
 batch_size: 128
 shuffle_all_dataloaders: False

hannah/conf/module/image_classifier.yaml

@@ -16,7 +16,7 @@
 ## See the License for the specific language governing permissions and
 ## limitations under the License.
 ##
-_target_: hannah.modules.ImageClassifierModule
+_target_: hannah.modules.vision.ImageClassifierModule
 num_workers: 0
 batch_size: 128
 shuffle_all_dataloaders: False

hannah/datasets/fake1d.py

@@ -26,44 +26,35 @@
 
 from ..utils.utils import extract_from_download_cache, list_all_files
 from .base import AbstractDataset, DatasetType
-from .vision.base import TorchvisionDatasetBase
 
 
-class Fake1dDataset(TorchvisionDatasetBase):
+class Fake1dDataset(AbstractDataset):
+    def __init__(self, config, size):
+        self.config = config
+        self.size = size
+
+        self.data = torch.randn((size, config["channels"], config["resolution"])).split(
+            1, 0
+        )
+        self.target = torch.randn(
+            (size, config.size), dtype=torch.int32, min=0, max=config["num_classes"]
+        )
+
     @classmethod
     def prepare(cls, config):
         pass
 
     @classmethod
     def splits(cls, config):
-        resolution = config.resolution
-        channels = config.channels
-
-        test_data = torchvision.datasets.FakeData(
-            size=128,
-            image_size=(channels, resolution),
-            num_classes=config.num_classes,
-        )
-        val_data = torchvision.datasets.FakeData(
-            size=128,
-            image_size=(channels, resolution),
-            num_classes=config.num_classes,
-        )
-        train_data = torchvision.datasets.FakeData(
-            size=512,
-            image_size=(channels, resolution),
-            num_classes=config.num_classes,
-        )
-
-        return cls(config, train_data), cls(config, val_data), cls(config, test_data)
+        return cls(config, size=128), cls(config, size=32), cls(config, size=32)
 
     def __getitem__(self, index):
-        data, target = self.dataset[index]
-        data = np.array(data).astype(np.float32) / 255
-        data = self.transform(image=data)["image"]
-        data = torch.squeeze(data)
+        data, target = self.data[index], self.target[index]
         return data, target
 
     @property
     def class_names(self):
         return [f"class{n}" for n in range(self.config.num_classes)]
+
+    def __len__(self):
+        return len(self.targets)

hannah/datasets/vision/base.py

@@ -18,20 +18,13 @@
 #
 import logging
 import re
-import tarfile
-from collections import Counter, namedtuple
-from typing import Dict, List, Optional
+from collections import Counter
+from typing import List
 
 import albumentations as A
 import cv2
 import numpy as np
-import pandas as pd
-import requests
-import torch
-import torchvision
 from albumentations.pytorch import ToTensorV2
-from omegaconf import DictConfig
-from sklearn.model_selection import train_test_split
 
 from ..base import AbstractDataset
 

hannah/datasets/vision/cifar.py

@@ -19,10 +19,8 @@
 import logging
 import os
 
-import albumentations as A
 import torch.utils.data as data
 import torchvision
-from albumentations.pytorch.transforms import ToTensorV2
 from torchvision import datasets
 
 from .base import TorchvisionDatasetBase

hannah/modules/__init__.py

@@ -27,14 +27,11 @@
     StreamClassifierModule,
 )
 from .object_detection import ObjectDetectionModule
-from .vision import AnomalyDetectionModule, ImageClassifierModule
 
 __all__ = [
-    "AnomalyDetectionModule",
     "CrossValidationStreamClassifierModule",
     "SpeechClassifierModule",
     "StreamClassifierModule",
-    "ImageClassifierModule",
     "AnomalyDetectionModule",
     "ObjectDetectionModule",
     "CartesianClassifierModule",