test out jupyter notebook

DeepForest.py

@@ -16,16 +16,19 @@ def parse_args():
     return known_args, unknown_args
 
 def convert_unknown_args_to_dict(unknown_args):
+    def set_nested_dict(d, keys, value):
+        for key in keys[:-1]:
+            d = d.setdefault(key, {})
+        d[keys[-1]] = value
+
     kwargs = {}
+    key = None
     for arg in unknown_args:
         if arg.startswith('--'):
-            key = arg.lstrip('--')
-            kwargs[key] = None
+            key = arg.lstrip('--').split('.')
+            set_nested_dict(kwargs, key, None)
         else:
-            if kwargs[key] is None:
-                kwargs[key] = arg
-            else:
-                kwargs[key] += f" {arg}"
+            set_nested_dict(kwargs, key, arg)
     return kwargs
 
 

deepforest_config.yml

@@ -45,7 +45,7 @@ train:
             eps: 1e-08
 
     # Print loss every n epochs
-    epochs: 1
+    epochs: 10
     # Useful debugging flag in pytorch lightning, set to True to get a single batch of training to test settings.
     fast_dev_run: False
     # pin images to GPU memory for fast training. This depends on GPU size and number of images.

docs/examples/Datasets.ipynb

@@ -251,6 +251,35 @@
     "m.config[\"train\"][\"csv_file\"] =\"<dummy file, existing dataloader>\"\n",
     "m.trainer.fit(m, train_loader)  "
    ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## Evaluation loader"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "from milliontrees.common.data_loaders import get_eval_loader\n",
+    "\n",
+    "test_dataset = dataset.get_subset(\"train\")\n",
+    "test_loader = get_eval_loader(\"standard\", test_dataset, batch_size=16)\n",
+    "\n",
+    "# Show one batch of the loader\n",
+    "for metadata, image, targets in test_loader:\n",
+    "    print(\"Targets is a list of dictionaries with the following keys: \", targets[0].keys())\n",
+    "    print(f\"Image shape: {image.shape}, Image type: {type(image)}\")\n",
+    "    print(f\"Annotation shape of the first image: {targets[0]['boxes'].shape}\")\n",
+    "    break  # Just show the first batch\n",
+    "\n",
+    "# Evaluate\n",
+    "dataset.eval(all_y_pred, all_y_true, all_metadata)"
+   ]
   }
  ],
  "metadata": {

tests/test_TreeBoxes.py

@@ -1,5 +1,5 @@
 from milliontrees.datasets.TreeBoxes import TreeBoxesDataset
-from milliontrees.common.data_loaders import get_train_loader
+from milliontrees.common.data_loaders import get_train_loader, get_eval_loader
 
 import torch
 import pytest
@@ -54,6 +54,31 @@ def test_get_train_dataloader(dataset, batch_size):
         assert len(metadata) == batch_size
         break
 
+def test_get_test_dataloader(dataset, batch_size):
+    dataset = TreeBoxesDataset(download=False, root_dir=dataset) 
+    test_dataset = dataset.get_subset("test")
+
+    for metadata, image, targets in test_dataset:
+        boxes, labels = targets["boxes"], targets["labels"]
+        assert image.shape == (100, 100, 3)
+        assert image.dtype == np.float32
+        assert image.min() >= 0.0 and image.max() <= 1.0
+        assert boxes.shape == (2, 4)
+        assert labels.shape == (2,)
+        assert metadata.shape == (2,2)
+        break
+
+    test_loader = get_eval_loader('standard', test_dataset, batch_size=batch_size)
+    for metadata, x, targets in test_loader:
+        y = targets[0]["boxes"]
+        assert torch.is_tensor(targets[0]["boxes"])
+        assert x.shape == (batch_size, 3, 448, 448)
+        assert x.dtype == torch.float32
+        assert x.min() >= 0.0 and x.max() <= 1.0
+        assert y.shape[1] == 4
+        assert len(metadata) == batch_size
+        break
+
 # Test structure with real annotation data to ensure format is correct
 # Do not run on github actions, long running.
 @pytest.mark.skipif(not on_hipergator, reason="Do not run on github actions")
@@ -93,5 +118,4 @@ def test_TreeBoxes_download(tmpdir):
         assert image.min() >= 0.0 and image.max() <= 1.0
         assert boxes.shape[1] == 4
         assert metadata.shape[1] == 1
-        break
-
+        break