From c19a0952fca1844e732e4e4288beb3358d74de0c Mon Sep 17 00:00:00 2001
From: zhang chen <chen.zhang@zilliz.com>
Date: Fri, 17 Dec 2021 15:38:49 +0800
Subject: [PATCH] fix dimension problem

---
 efficientnet_image_embedding.py | 5 ++---
 pytorch/model.py                | 5 ++++-
 2 files changed, 6 insertions(+), 4 deletions(-)

diff --git a/efficientnet_image_embedding.py b/efficientnet_image_embedding.py
index bb62091..418ca4f 100644
--- a/efficientnet_image_embedding.py
+++ b/efficientnet_image_embedding.py
@@ -43,8 +43,7 @@ class EfficientnetImageEmbedding(Operator):
                                         transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225]), ])
 
     def __call__(self, img_path: str) -> NamedTuple('Outputs', [('feature_vector', numpy.ndarray)]):
-        Outputs = NamedTuple('Outputs', [('embedding', torch.Tensor)])
         img = self.tfms(Image.open(img_path)).unsqueeze(0)
-        features = self.model._model.extract_features(img)
+        features = self.model(img)
         Outputs = NamedTuple('Outputs', [('feature_vector', numpy.ndarray)])
-        return Outputs(features.flatten().detach().numpy())
+        return Outputs(features)
diff --git a/pytorch/model.py b/pytorch/model.py
index d973dac..507eee6 100644
--- a/pytorch/model.py
+++ b/pytorch/model.py
@@ -27,10 +27,13 @@ class Model():
     def __init__(self, model_name: str, weights_path: str):
         super().__init__()
         self._model = EfficientNet.from_pretrained(model_name=model_name, weights_path=weights_path)
+        self._avg_pooling = torch.nn.AdaptiveAvgPool2d((1, 1))
         self._model.eval()
 
     def __call__(self, img_tensor: torch.Tensor):
-        return self._model(img_tensor).detach().numpy()
+        features = self._model.extract_features(img_tensor)
+        channels = features.shape[0]
+        return self._avg_pooling(features).view(channels, -1).detach().numpy()
 
     def train(self):
         """