[Image/Mask loaders] Allow image recycle

So when we have multiple instances of the same image we load just once

src/nodes/helpers.py

@@ -147,13 +147,16 @@ def get_image_preview_info(file_name, where="input"):
     return {"filename": fname, "subfolder": dname, "type": where}
 
 
-def load_one_image(file_name, disp_name, embed_transparency):
+def load_one_image(file_name, disp_name, embed_transparency, known=None):
     if file_name is None:
         return (empty_image(b=1, c=3), empty_image(b=1), file_name)
     if not os.path.isabs(file_name):
         file_name = os.path.join(get_input_directory(), file_name)
     if not os.path.exists(file_name):
         raise ValueError(f"File '{file_name}' not found")
+    if known is not None and file_name in known:
+        logger.debug('Recycling '+file_name)
+        return (known[file_name][0], known[file_name][1], file_name)
 
     try:
         if has_load_image:
@@ -182,6 +185,8 @@ def load_one_image(file_name, disp_name, embed_transparency):
             # Concatenate image and mask into (b, h, w, 4)
             image_with_alpha = torch.cat([image, 1.0 - mask], dim=-1)
             result = (image_with_alpha, mask, file_name)
+        if known is not None:
+            known[file_name] = (result[0], result[1])
         return result
 
     except Exception as e:
@@ -191,13 +196,16 @@ def load_one_image(file_name, disp_name, embed_transparency):
                       "It may be corrupt or in an unsupported format.") from e
 
 
-def load_one_mask(file_name, disp_name, channel='red'):
+def load_one_mask(file_name, disp_name, channel='red', known=None):
     if file_name is None:
         return (empty_image(b=1), file_name)
     if not os.path.isabs(file_name):
         file_name = os.path.join(get_input_directory(), file_name)
     if not os.path.exists(file_name):
         raise ValueError(f"File '{file_name}' not found")
+    if known is not None and file_name in known:
+        logger.debug('Recycling '+file_name)
+        return (known[file_name], file_name)
 
     try:
         if has_load_image:
@@ -210,6 +218,8 @@ def load_one_mask(file_name, disp_name, channel='red'):
 
         # Call the method and return its result directly
         result = loader_instance.load_image(file_name, channel)
+        if known is not None:
+            known[file_name] = result[0]
         return (result[0], file_name)
 
     except Exception as e:
@@ -241,6 +251,8 @@ def load_images_wrapper(file_names, embed_transparency=False, disp_names=None, s
 
     imgs = []
     masks = []
+    known_imgs = {}
+    known_masks = {}
     is_mask = channel is not None
     used_file_names = []
     all_preview_imgs = []
@@ -261,9 +273,9 @@ def load_images_wrapper(file_names, embed_transparency=False, disp_names=None, s
 
                 if is_mask:
                     # A mask
-                    mask, file_name = load_one_mask(file_name, disp_name, channel)
+                    mask, file_name = load_one_mask(file_name, disp_name, channel, known=known_masks)
                 else:
-                    img, mask, file_name = load_one_image(file_name, disp_name, embed_transparency)
+                    img, mask, file_name = load_one_image(file_name, disp_name, embed_transparency, known=known_imgs)
 
                     max_w = max(max_w, img.shape[2])
                     max_h = max(max_h, img.shape[1])
@@ -279,7 +291,7 @@ def load_images_wrapper(file_names, embed_transparency=False, disp_names=None, s
                     all_preview_imgs.append(get_image_preview_info(file_name))
 
             for j in range(len(imgs_batch)):
-                if channel is None:
+                if not is_mask:
                     img = imgs_batch[j]
                     H, W = img.shape[1:3]
                     if H != max_h or W != max_w:
@@ -292,7 +304,7 @@ def load_images_wrapper(file_names, embed_transparency=False, disp_names=None, s
                     logger.debug(f"Upscaling mask to fit batch: {W}x{H} -> {max_mw}x{max_mh}")
                     masks_batch[j] = upscale_comfy(mask, max_mw, max_mh, "bicubic")
 
-            if channel is None:
+            if not is_mask:
                 imgs.append(torch.cat(imgs_batch))
             masks.append(torch.cat(masks_batch))
         else:
@@ -301,9 +313,9 @@ def load_images_wrapper(file_names, embed_transparency=False, disp_names=None, s
 
             if is_mask:
                 # A mask
-                mask, file_name = load_one_mask(file_name, disp_names[i], channel)
+                mask, file_name = load_one_mask(file_name, disp_names[i], channel, known=known_masks)
             else:
-                img, mask, file_name = load_one_image(file_name, disp_names[i], embed_transparency)
+                img, mask, file_name = load_one_image(file_name, disp_names[i], embed_transparency, known=known_imgs)
                 imgs.append(img)
 
             masks.append(mask)

src/nodes/nodes_img.py

@@ -3238,6 +3238,7 @@ class ImageMemoryEstimator:
         BYTES_PER_PIXEL = 4 * channels
         logger.debug(f"BYTES_PER_PIXEL {BYTES_PER_PIXEL}")
 
+        known = set()
         for fname in file_name:
             if fname is None:
                 continue
@@ -3248,6 +3249,9 @@ class ImageMemoryEstimator:
                 logger.warning(f"Could not find file: {fname}")
                 continue
 
+            if file_path in known:
+                continue
+
             try:
                 # PIL.Image.open is lazy; it reads metadata without loading pixel data
                 with Image.open(file_path) as img:
@@ -3264,6 +3268,7 @@ class ImageMemoryEstimator:
                         max_dim = (w, h)
 
                     valid_count += 1
+                    known.add(file_path)
             except Exception as e:
                 logger.error(f"Failed to read metadata for {fname}: {e}")