Add image perspective editing and related tools

- Implemented the Image Perspective tool for editing image corners in Blender.
- Created a new operator `EditImagePerspective` to handle the perspective transformation.
- Added a preview feature for the perspective adjustment.
- Introduced the `ImagePolylineTool` for editing image masks with polylines.
- Developed backend support for Real-ESRGAN model inference, including ONNX integration.
- Added functionality for image masking and background removal.
- Created unit tests for the Real-ESRGAN ONNX model to ensure correct preprocessing and inference.

tests/test_server.py

@@ -79,7 +79,14 @@ class ServerTest(unittest.TestCase):
             snapshot = manager.snapshot()
         self.assertEqual(
             snapshot,
-            {"loaded": {"ben2": "", "da3": "", "moge2": ""}},
+            {
+                "loaded": {
+                    "ben2": "",
+                    "da3": "",
+                    "moge2": "",
+                    "realesrgan": "",
+                }
+            },
         )
         self.assertIn("server.model_catalog", sys.modules)
         self.assertNotIn("model_catalog", sys.modules)
@@ -94,6 +101,8 @@ class ServerTest(unittest.TestCase):
         cache.da3_key = None
         cache.moge_model = None
         cache.moge_key = None
+        cache.realesrgan_model = None
+        cache.realesrgan_key = None
         return cache
 
     def create_server(self):
@@ -184,6 +193,25 @@ class ServerTest(unittest.TestCase):
         self.assertIsNone(cache.da3_model)
         self.assertIsNotNone(cache.ben2_model)
 
+    def test_realesrgan_model_is_reused_for_matching_runtime_key(self):
+        loads = []
+        realesrgan = ModuleType("server.models.onnx_realesrgan")
+        realesrgan.create_session = lambda model_dir, device: (
+            loads.append((model_dir, device)) or object()
+        )
+        cache = self.create_cache()
+
+        with patch.dict(
+            sys.modules,
+            {"server.models.onnx_realesrgan": realesrgan},
+        ):
+            first = cache.get_realesrgan("model", "cpu")
+            second = cache.get_realesrgan("model", "cpu")
+
+        self.assertIs(first, second)
+        self.assertEqual(loads, [("model", "cpu")])
+        self.assertEqual(cache.snapshot()["realesrgan"], "model")
+
     def test_dedicated_server_runs_only_one_job_at_a_time(self):
         server = self.server_module.JobServer("Test", storage_root=self.storage_root)
         first_started = threading.Event()
@@ -220,7 +248,7 @@ class ServerTest(unittest.TestCase):
                 "download-model",
                 "download-required-models",
                 "analyze-balloon",
-                "extract-image-background",
+                "refine-image-mask",
                 "remove-background",
                 "extract-frames",
                 "generate-depth",
@@ -228,6 +256,7 @@ class ServerTest(unittest.TestCase):
                 "debug-da3",
                 "debug-moge2",
                 "debug-ben2",
+                "debug-realesrgan-general-x4v3",
             },
         )
 
@@ -401,16 +430,16 @@ class ServerTest(unittest.TestCase):
         self.assertEqual(tuple(result[0, 0]), (20, 40, 60, 64))
         self.assertEqual(tuple(result[3, 3]), (0, 0, 0, 0))
 
-    def test_image_extract_canvas_clears_rgba_outside_the_selection(self):
+    def test_image_mask_canvas_clears_rgba_outside_the_selection(self):
         import numpy as np
         from PIL import Image
 
-        image_extract = sys.modules["server.jobs.image_extract"]
+        image_mask = sys.modules["server.jobs.image_mask"]
         source_path = self.storage_root / "source.png"
         pixels = np.full((2, 2, 4), (20, 40, 60, 255), dtype=np.uint8)
         Image.fromarray(pixels, mode="RGBA").save(source_path)
 
-        canvas, _selection, bounds = image_extract.create_selection_canvas(
+        canvas, _selection, bounds = image_mask.create_selection_canvas(
             source_path,
             ((-1, -1), (2, -1), (-1, 2)),
         )
@@ -421,25 +450,22 @@ class ServerTest(unittest.TestCase):
         self.assertEqual(tuple(result[2, 2]), (0, 0, 0, 0))
         self.assertEqual(tuple(result[1, 1]), (20, 40, 60, 255))
 
-    def test_image_extract_crops_to_alpha_with_pixel_padding(self):
+    def test_image_mask_crops_tightly_to_alpha(self):
         import numpy as np
         from PIL import Image
 
-        image_extract = sys.modules["server.jobs.image_extract"]
+        image_mask = sys.modules["server.jobs.image_mask"]
         pixels = np.zeros((4, 5, 4), dtype=np.uint8)
         pixels[1:3, 2:4] = (20, 40, 60, 255)
-        result, bounds = image_extract.crop_to_alpha(
+        result, bounds = image_mask.crop_to_alpha(
             Image.fromarray(pixels, mode="RGBA"),
             (10, 20, 15, 24),
-            padding=1,
         )
         result = np.asarray(result)
 
-        self.assertEqual(bounds, (11, 20, 15, 24))
-        self.assertEqual(result.shape, (4, 4, 4))
-        self.assertTrue(np.all(result[[0, -1], :, 3] == 0))
-        self.assertTrue(np.all(result[:, [0, -1], 3] == 0))
-        self.assertTrue(np.all(result[1:3, 1:3, 3] == 255))
+        self.assertEqual(bounds, (12, 21, 14, 23))
+        self.assertEqual(result.shape, (2, 2, 4))
+        self.assertTrue(np.all(result[:, :, 3] == 255))
 
     def test_balloon_job_sends_the_masked_crop_to_ben2(self):
         import numpy as np