time to fine tune again. Not sure if I should use previous fine tune …

…or just start over.

.gitignore

@@ -163,3 +163,5 @@ cython_debug/
 #  and can be added to the global gitignore or merged into this file.  For a more nuclear
 #  option (not recommended) you can uncomment the following to ignore the entire idea folder.
 #.idea/
+/predictions_round2/
+/sp_photos_yolo11/

10_final_predictions.py

@@ -0,0 +1,10 @@
+import fiftyone as fo
+import os
+import tempfile
+import torch
+from ultralytics import YOLO
+
+if __name__ == '__main__':
+    print("starting")
+
+    print("finished")

6_fine_tuning.py

@@ -3,9 +3,7 @@
 import tempfile
 import torch
 from ultralytics import YOLO
-# from ultralytics import settings
 import fiftyone as fo
-# import wandb
 from wandb.sdk.verify.verify import PROJECT_NAME
 
 """
@@ -21,20 +19,18 @@
 Requires `ultralytics` and `fiftyone>=0.25.0` to be installed.
 ---------------
 Steve says: 
-"I was going to to start with the imgsz =640x640 by resizing to 860x860 and then crop to 640x640. I can't
+"//I was going to to start with the imgsz =640x640 by resizing to 860x860 and then crop to 640x640//. I can't
 do this because the image size below is just a magic number we set with the Yolo library. It says it 
 will just resize to that size but doesn't tell us how. So instead, to start with, I am just setting this to 640.
-I am also setting the model size to nano, but will bump it up over time.
+I am also setting the model size to nano, but will bump it up over time. Settled on X - it doesn't take much longer and does better
 """
 
-
 DATASET_NAME = 'labeled_dataset'
 DEFAULT_MODEL_SIZE = "x"
 DEFAULT_IMAGE_SIZE = 704
 DEFAULT_EPOCHS = 12
 PROJECT_NAME = 'sp_photos_yolo11'
 
-# wandb.require("core")
 
 
 def get_torch_device():

8_cleaning_ground_truth_round2.py

@@ -1,16 +1,21 @@
 import fiftyone as fo
 import argparse
+from fiftyone import ViewField as F
 
 """ 
 This script needs to accept arguments because we don't want to do all the tasks every time we run it.
 * at the end means it happens in the app, not here in code
 So what we do is:
 0. load in the new dataset - "second_play_photos"
-1. Create a new field for ground truth
-2. One at a time, go through each of the categories and just view that category
-3. If there is a mislabel, then fix it in the ground truth field. 
-4. Once all the images have been fixed for that category, head on to the next category
-5. When finished all the categories, anything that has a blank ground truth field should have the ground
+1. * Go through the samples and if the prediction is incorrect then add a tag with the correct field 
+2. After first pass look at all teh untagged images and make sure they are correct:
+```
+from fiftyone import ViewField as F
+view = dataset.match(F("tags") == [])
+```
+
+
+5. When finished all the images, anything that has a blank tag should have the ground
 truth field set = to the prediction field. 
 Then we are ready to do our final training/test
 Then final predictions
@@ -29,28 +34,44 @@
     dataset = fo.load_dataset(DATASET_NAME)
 
     parser = argparse.ArgumentParser()
-    parser.add_argument("-g", help="specify to create ground truth field", action="store_true")
+    # parser.add_argument("-g", help="specify to create ground truth field", action="store_true")
     parser.add_argument("-p", help="specify to move correct predictions to ground truth field",
                         action="store_true")
     args = parser.parse_args()
 
-    if args.g:
-        print("make the field")
+    # if args.g:
+    #     print("make the field")
+    #     if dataset.has_field("ground_truth"):
+    #         dataset.delete_sample_field("ground_truth")
+    #     dataset.add_sample_field(
+    #         "ground_truth",
+    #         fo.EmbeddedDocumentField,
+    #         embedded_doc_type=fo.Classification,
+    #     )
+    #     dataset.save()
+    if args.p:
+        # Add ground_truth label to dataset
         if dataset.has_field("ground_truth"):
             dataset.delete_sample_field("ground_truth")
-        dataset.add_sample_field(
+        # Must initialize new `Label` values via `set_values()`
+        # This basically creates a Classification for each sample in the data set which just contains an id
+        dataset.set_values(
             "ground_truth",
-            fo.EmbeddedDocumentField,
-            embedded_doc_type=fo.Classification,
+            [fo.Classification() for _ in range(len(dataset))],
         )
-        dataset.save()
-    elif args.p:
-        for sample in dataset.iter_samples(autosave=True):
-            if sample["ground_truth"] is None:
-                sample["ground_truth"] = sample["prediction"]
 
+        # Expression that grabs first tag if one exists, else falls back to the predicted label
+        # We need an expression for set_field
+        # This is basically saying "Test if tags is not empty, if that is true then return the first tag, else return the prediction label
+        label_expr = (F("$tags") != []).if_else(
+            F("$tags")[0],
+            F("$prediction.label"),
+        )
+
+        # Apply changes
+        dataset.set_field("ground_truth.label", label_expr).save()
     else:
-        print("you need to specify either -g or -p")
+        # print("you need to specify either -g or -p")
+        print("you need to specify  -p")
 
     print("done")
-

9_final_fine_tuning.py

@@ -0,0 +1,29 @@
+import fiftyone as fo
+import os
+import tempfile
+import torch
+from ultralytics import YOLO
+
+FIRST_TRAINING = "low_quality_first_labeled_dataset"
+SECOND_TRAINING = ""
+
+DATASET_NAME = 'labeled_dataset'
+DEFAULT_MODEL_SIZE = "x"
+DEFAULT_IMAGE_SIZE = 704
+DEFAULT_EPOCHS = 12
+PROJECT_NAME = 'sp_photos_yolo11'
+
+
+
+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")
+
+if __name__ == '__main__':
+    print("starting")
+
+    print("finished")

README.md

@@ -1,3 +1,11 @@
+@@TODO I need to rename the datasets, and their references in code, to make more sense
+The way to rename a dataset is
+```python
+import fiftyone as fo
+dataset = fo.load_dataset("foo")
+dataset.name = "footwo"
+```
+
 # voxel-photo-album
 
 Here are the general steps we want to accomplish