fixing image input processing

app.py

@@ -8,6 +8,8 @@ import os
 feature_extractor = SegformerFeatureExtractor.from_pretrained("nvidia/segformer-b5-finetuned-cityscapes-1024-1024")
 model = SegformerForSemanticSegmentation.from_pretrained("nvidia/segformer-b5-finetuned-cityscapes-1024-1024")
 
+# https://github.com/NielsRogge/Transformers-Tutorials/blob/master/SegFormer/Segformer_inference_notebook.ipynb
+
 def cityscapes_palette():
     """Cityscapes palette for external use."""
     return [[128, 64, 128], [244, 35, 232], [70, 70, 70], [102, 102, 156],
@@ -42,13 +44,19 @@ def annotation(image:ImageDraw, color_seg:np.array):
         sub_square_seg = reduced_seg[ y:y+step_size, x:x+step_size]
         # print(f"{sub_square_seg.shape=}, {sub_square_seg.sum()}")
         
-        if (sub_square_seg.sum() > 1000000): 
+        if (sub_square_seg.sum() > 100000): 
             print("light found at square ", x, y)
             draw.rectangle([(x, y), (x + step_size, y + step_size)], outline=128, width=3)
 
-def call(image: Image):
-    resized_image = original_image.resize((1024,1024))
+def call(image): #nparray
+
+    resized = Image.fromarray(image).resize((1024,1024))
+    resized_image = np.array(resized)
     print(f"{np.array(resized_image).shape=}") # 1024, 1024, 3
+
+    # resized_image = Image.fromarray(resized_image_np)
+    # print(f"{resized_image=}")
+
     inputs = feature_extractor(images=resized_image, return_tensors="pt")
 
     outputs  = model(**inputs)
@@ -58,7 +66,7 @@ def call(image: Image):
     # First, rescale logits to original image size
     interpolated_logits =  nn.functional.interpolate(
         outputs.logits,
-        size=resized_image.size[::-1], # (height, width)
+        size=[1024, 1024], #resized_image.size[::-1], # (height, width)
         mode='bilinear',
         align_corners=False)
     print(f"{interpolated_logits.shape=}, {outputs.logits.shape=}") # 1, 19, 1024, 1024
@@ -85,8 +93,8 @@ def call(image: Image):
 
     return out_im_file
 
-original_image = Image.open("./examples/1.jpg")
-print(f"{np.array(original_image).shape=}") # eg 729, 1000, 3
+# original_image = Image.open("./examples/1.jpg")
+# print(f"{np.array(original_image).shape=}") # eg 729, 1000, 3
 
 # out = call(original_image)
 # out.save("out2.jpeg")
@@ -101,7 +109,11 @@ iface = gr.Interface(fn=call,
                      description=description, 
                      examples=[
                        os.path.join(os.path.dirname(__file__), "examples/1.jpg"),
-                       os.path.join(os.path.dirname(__file__), "examples/2.jpg")
+                       os.path.join(os.path.dirname(__file__), "examples/2.jpg"),
+                       os.path.join(os.path.dirname(__file__), "examples/3.jpg"),
+                       os.path.join(os.path.dirname(__file__), "examples/4.jpg"),
+                       os.path.join(os.path.dirname(__file__), "examples/5.jpg"),
+                       os.path.join(os.path.dirname(__file__), "examples/6.jpg"),
                      ],
                      thumbnail="thumbnail.webp")
 iface.launch()