Help from Jacob

4_generate_labels.py

@@ -3,9 +3,13 @@
 import fiftyone as fo
 import fiftyone.zoo as foz
 
-
 if __name__ == '__main__':
     dataset = fo.load_dataset("play_photos")
+
+    # Clean up from previous runs
+    if "labeled_dataset" in fo.list_datasets():
+        fo.delete_dataset("labeled_dataset")
+
     clip = foz.load_zoo_model(
         "clip-vit-base32-torch",
         text_prompt="A photo of a",
@@ -14,14 +18,20 @@
     # alexnet = foz.load_zoo_model("alexnet-imagenet-torch")
     # dense201 = foz.load_zoo_model("densenet201-imagenet-torch")
     # fasterrcnn = foz.load_zoo_model("faster-rcnn-resnet50-fpn-coco-torch")
-    # yoloseg = foz.load_zoo_model("yolo11x-seg-coco-torch")
+    #yoloseg = foz.load_zoo_model("yolo11x-seg-coco-torch")
 
-    dataset.apply_model(clip, label_field="default_prediction")
-    dataset.set_values("ground_truth", fo.Classification(label=str(dataset.values("default_prediction.label"))))
+    dataset.apply_model(clip, label_field="prediction")
     # dataset.apply_model(dense201, label_field="dense201")
     # dataset.apply_model(alexnet, label_field="alexnet")
     # dataset.apply_model(fasterrcnn, label_field="faster_rcnn")
-    # dataset.apply_model(yoloseg, label_field="yolo_seg")
+    #dataset.apply_model(yoloseg, label_field="yolo_seg")
+
+    #Alright time to make our dataset with cleaned labels
+    labeled_dataset = dataset.clone(name="labeled_dataset", persistent=True)
+    labeled_dataset.rename_sample_field("prediction", "ground_truth")
+    labeled_dataset.set_field("ground_truth.detections.confidence", None).save()
+
+    # Now time to go to 5_clean_ground_truth
 
     session = fo.launch_app(dataset)
     session.wait()

5_clean_ground_truth.py

@@ -0,0 +1,4 @@
+import fiftyone as fo
+
+if __name__ == '__main__':
+    fo.load_dataset("labeled_dataset")

6_fine_tuning.py

@@ -0,0 +1,119 @@
+import wandb
+from plotly.validators.layout.slider import StepsValidator
+"""
+
+Ultralytics YOLOv8*-cls model training script
+for generating confidence-based noise labels for a dataset.
+
+| Copyright 2017-2024, Voxel51, Inc.
+| `voxel51.com <https://voxel51.com/>`_
+|
+
+Requires `ultralytics` and `fiftyone>=0.25.0` to be installed.
+"""
+import argparse
+import os
+import tempfile
+import torch
+from ultralytics import YOLO
+from ultralytics import settings
+import fiftyone as fo
+
+DEFAULT_MODEL_SIZE = "s"
+DEFAULT_IMAGE_SIZE = 128
+DEFAULT_EPOCHS = 10
+
+wandb.require("core")
+
+
+def get_torch_device():
+    if torch.cuda.is_available():
+        return torch.device("cuda")
+    elif torch.backends.mps.is_available():
+        return torch.device("mps")
+    else:
+        return torch.device("cpu")
+
+
+def train_classifier(
+        dataset_name=None,
+        model_size=DEFAULT_MODEL_SIZE,
+        image_size=DEFAULT_IMAGE_SIZE,
+        epochs=DEFAULT_EPOCHS,
+        project_name="mislabel_confidence_noise",
+        gt_field="ground_truth",
+        train_split=None,
+        test_split=None,
+        **kwargs
+):
+
+    settings.update({"wandb": False})
+    if dataset_name:
+        dataset = fo.load_dataset(dataset_name)
+        train = dataset.match_tags("train")
+        test = dataset.match_tags("test")
+    else:
+        train = train_split
+        test = test_split
+
+    if model_size is None:
+        model_size = "s"
+    elif model_size not in ["n", "s", "m", "l", "x"]:
+        raise ValueError("model_size must be one of ['n', 's', 'm', 'l', 'x']")
+
+    splits_dict = {
+        "train": train,
+        "val": test,
+        "test": test,
+    }
+
+    data_dir = tempfile.mkdtemp()
+
+    for key, split in splits_dict.items():
+        split_dir = os.path.join(data_dir, key)
+        os.makedirs(split_dir)
+        split.export(
+            export_dir=split_dir,
+            dataset_type=fo.types.ImageClassificationDirectoryTree,
+            label_field=gt_field,
+            export_media="symlink",
+        )
+
+    # Load a pre-trained YOLOv8 model for classification
+    model = YOLO(f"yolov8{model_size}-cls.pt")
+
+    # Train the model
+    model.train(
+        data=data_dir,  # Path to the dataset
+        epochs=epochs,  # Number of epochs
+        imgsz=image_size,  # Image size
+        device=get_torch_device(),
+        project=project_name,
+    )
+
+    return model
+
+
+def main():
+    if fo.__version__ < "0.25.0":
+        raise ValueError("Please upgrade to the latest version of FiftyOne")
+
+    # parser = argparse.ArgumentParser()
+    # parser.add_argument("--dataset_name", type=str, required=True)
+    # parser.add_argument("--model_size", type=str, default=None)
+    # parser.add_argument("--image_size", type=int, default=128)
+    # parser.add_argument("--epochs", type=int, default=10)
+    # parser.add_argument("--project_name", type=str, default="mislabel_confidence_noise")
+    # args = parser.parse_args()
+
+    train_classifier(
+        dataset_name=args.dataset_name,
+        model_size=args.model_size,
+        image_size=args.image_size,
+        epochs=args.epochs,
+        project_name=args.project_name,
+    )
+
+
+if __name__ == "__main__":
+    main()