Refactor image processing in batch to handle grayscale images and unify prompts for model input

detect-objects.py

@@ -224,38 +224,29 @@ def process_batch(
     for img in images:
         if isinstance(img, str):
             img = Image.open(img)
-        if img.mode == "L" or img.mode != "RGB":
+        if img.mode == "L":
+            img = img.convert("RGB")
+        elif img.mode != "RGB":
             img = img.convert("RGB")
         pil_images.append(img)
 
-    # Store original sizes for post-processing
-    original_sizes = [(img.height, img.width) for img in pil_images]
-
     # Process batch through model
     try:
         inputs = processor(
             images=pil_images,
-            text=class_name,  # Single class name as prompt
+            text=[class_name] * len(pil_images),  # Same prompt for all images
             return_tensors="pt",
-        )
-        # Move to device and convert to model's dtype
-        inputs = {
-            k: v.to(
-                model.device,
-                dtype=model.dtype if v.dtype.is_floating_point else v.dtype,
-            )
-            for k, v in inputs.items()
-        }
+        ).to(model.device)
 
         with torch.no_grad():
             outputs = model(**inputs)
 
-        # Post-process outputs
+        # Post-process outputs using original_sizes from processor
         results = processor.post_process_instance_segmentation(
             outputs,
             threshold=confidence_threshold,
             mask_threshold=mask_threshold,
-            target_sizes=original_sizes,
+            target_sizes=inputs.get("original_sizes").tolist(),
         )
 
     except Exception as e: