From 98a10ca0e76a085c21788a94945136e75b140f5c Mon Sep 17 00:00:00 2001
From: ASHWIN UNNIKRISHNAN <unnikrishnan.a@northeastern.edu>
Date: Sun, 25 Feb 2024 22:31:30 -0500
Subject: [PATCH 1/3] Update inference.py

Adding additional function transform_image.. so that users can use a loaded RGB image and transform it to the expected format of grounding dino prediction.
---
 groundingdino/util/inference.py | 34 +++++++++++++++++++++++++++++++++
 1 file changed, 34 insertions(+)

diff --git a/groundingdino/util/inference.py b/groundingdino/util/inference.py
index 58528ed..88f9ec0 100644
--- a/groundingdino/util/inference.py
+++ b/groundingdino/util/inference.py
@@ -37,6 +37,17 @@ def load_model(model_config_path: str, model_checkpoint_path: str, device: str =
 
 
 def load_image(image_path: str) -> Tuple[np.array, torch.Tensor]:
+    """
+    Load an image and apply transformations.
+    This function takes the path to an image file, loads the image, and applies a series of transformations to it.
+    The transformations include resizing the image, converting it to a tensor, and normalizing its pixel values.
+
+    Parameters:
+    image_path (str): The path to the image file.
+
+    Returns:
+    Tuple[np.array, torch.Tensor]: A tuple containing the original image as a NumPy array and the transformed image as a PyTorch tensor.
+    """
     transform = T.Compose(
         [
             T.RandomResize([800], max_size=1333),
@@ -49,6 +60,29 @@ def load_image(image_path: str) -> Tuple[np.array, torch.Tensor]:
     image_transformed, _ = transform(image_source, None)
     return image, image_transformed
 
+def transform_image(t: PIL.Image.Image) -> Tuple[np.array, torch.Tensor]:
+    """
+    Transform an RGB image and convert it to a tensor.
+    
+    This function takes a PIL Image, applies a series of transformations to it, and returns the original and transformed images.
+    The transformations include resizing the image, converting it to a tensor, and normalizing its pixel values.
+
+    Parameters:
+    img (PIL.Image.Image): The input image.
+
+    Returns:
+    Tuple[np.array, torch.Tensor]: A tuple containing the original image as a NumPy array and the transformed image as a PyTorch tensor.
+    """
+    transform = T.Compose(
+        [
+            T.RandomResize([800], max_size=1333),
+            T.ToTensor(),
+            T.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225]),
+        ]
+    )
+    image = np.asarray(t)
+    image_transformed, _ = transform(t, None)
+    return image, image_transformed
 
 def predict(
         model,

From da66253d1baf38109018c928bf0ba19571274d42 Mon Sep 17 00:00:00 2001
From: ASHWIN UNNIKRISHNAN <unnikrishnan.a@northeastern.edu>
Date: Sun, 25 Feb 2024 23:11:49 -0500
Subject: [PATCH 2/3] Update inference.py

---
 groundingdino/util/inference.py | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/groundingdino/util/inference.py b/groundingdino/util/inference.py
index 88f9ec0..7dd5fb1 100644
--- a/groundingdino/util/inference.py
+++ b/groundingdino/util/inference.py
@@ -60,7 +60,7 @@ def load_image(image_path: str) -> Tuple[np.array, torch.Tensor]:
     image_transformed, _ = transform(image_source, None)
     return image, image_transformed
 
-def transform_image(t: PIL.Image.Image) -> Tuple[np.array, torch.Tensor]:
+def transform_image(PIL_image: PIL.Image.Image) -> Tuple[np.array, torch.Tensor]:
     """
     Transform an RGB image and convert it to a tensor.
     
@@ -68,7 +68,7 @@ def transform_image(t: PIL.Image.Image) -> Tuple[np.array, torch.Tensor]:
     The transformations include resizing the image, converting it to a tensor, and normalizing its pixel values.
 
     Parameters:
-    img (PIL.Image.Image): The input image.
+    PIL_image (PIL.Image.Image): The input image.
 
     Returns:
     Tuple[np.array, torch.Tensor]: A tuple containing the original image as a NumPy array and the transformed image as a PyTorch tensor.
@@ -80,7 +80,7 @@ def transform_image(t: PIL.Image.Image) -> Tuple[np.array, torch.Tensor]:
             T.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225]),
         ]
     )
-    image = np.asarray(t)
+    image = np.asarray(PIL_image)
     image_transformed, _ = transform(t, None)
     return image, image_transformed
 

From c176fd543bc28f0c7d8875867c29901c2f9133a7 Mon Sep 17 00:00:00 2001
From: ASHWIN UNNIKRISHNAN <unnikrishnan.a@northeastern.edu>
Date: Sun, 25 Feb 2024 23:14:54 -0500
Subject: [PATCH 3/3] Update inference.py

---
 groundingdino/util/inference.py | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/groundingdino/util/inference.py b/groundingdino/util/inference.py
index 7dd5fb1..20353db 100644
--- a/groundingdino/util/inference.py
+++ b/groundingdino/util/inference.py
@@ -81,7 +81,7 @@ def transform_image(PIL_image: PIL.Image.Image) -> Tuple[np.array, torch.Tensor]
         ]
     )
     image = np.asarray(PIL_image)
-    image_transformed, _ = transform(t, None)
+    image_transformed, _ = transform(PIL_image, None)
     return image, image_transformed
 
 def predict(