Upload processor

image_processor.py

@@ -23,7 +23,7 @@ from transformers.image_utils import (
 from transformers.utils import is_torch_tensor
 
 
-class FaceSegformerImageProcessor(BaseImageProcessor):
+class FaceNetImageProcessor(BaseImageProcessor):
     def __init__(self, **kwargs):
         super().__init__(**kwargs)
         self.image_size = kwargs.get("image_size", (224, 224))
@@ -204,7 +204,7 @@ class FaceSegformerImageProcessor(BaseImageProcessor):
         data = {"pixel_values": images}
         return BatchFeature(data=data, tensor_type="pt")
 
-    # Copied from transformers.models.segformer.image_processing_segformer.SegformerImageProcessor.post_process_semantic_segmentation
+    # Copied from transformers.models.beit.image_processing_beit.BeitImageProcessor.post_process_semantic_segmentation with Beit->Segformer
     def post_process_semantic_segmentation(
         self, outputs, target_sizes: List[Tuple] = None
     ):
@@ -255,3 +255,36 @@ class FaceSegformerImageProcessor(BaseImageProcessor):
 
         return semantic_segmentation
 
+    # def post_process_instance_segmentation(
+    #     self,
+    #     outputs,
+    #     target_sizes: List[Tuple] = None,
+    #     **kwargs
+    # ) -> np.ndarray:
+    #     logits = outputs.logits
+    #     if target_sizes is not None:
+    #         if len(logits) != len(target_sizes):
+    #             raise ValueError(
+    #                 "Make sure that you pass in as many target sizes as the batch dimension of the logits"
+    #             )
+    #         # use target sizes to resize logits
+    #         resized_masks = []
+    #         for idx in range(len(logits)):
+    #             resized_mask = torch.nn.functional.interpolate(
+    #                 logits[idx].unsqueeze(dim=0),
+    #                 size=target_sizes[idx],
+    #                 mode="bilinear",
+    #                 align_corners=False,
+    #             )
+    #             resized_masks.append(
+    #                 resized_mask[0].softmax(dim=0).argmax(dim=0).to("cpu").numpy()
+    #             )
+
+    #         predicted_masks = np.array(resized_masks)
+
+    #     else:
+    #         predicted_masks = (
+    #             torch.argmax(torch.softmax(logits, dim=1), axis=1).to("cpu").numpy()
+    #         )
+
+    #     return predicted_masks

preprocessor_config.json

@@ -1,9 +1,9 @@
 {
   "auto_map": {
-    "AutoImageProcessor": "image_processor.FaceSegformerImageProcessor"
+    "AutoImageProcessor": "image_processor.FaceNetImageProcessor"
   },
   "data_format": "channels_first",
-  "image_processor_type": "FaceSegformerImageProcessor",
+  "image_processor_type": "FaceNetImageProcessor",
   "image_size": [
     224,
     224