change onnx to NHWC

FPN_int.onnx → FPN_int_NHWC.onnx

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:4a172eb921a5119875bd12561e658450ca4dae95c4aa2ea350dfd603cd27f14a
-size 45595505
+oid sha256:d1cca42fe8ac3429c55aa1a28a3ecdf08cd279b6193e6487d9b18fa28cfc2c7e
+size 45595686

infer_onnx.py

@@ -37,8 +37,8 @@ if __name__ == "__main__":
     onnx_path = args.onnx_path
     input_img = build_img(args)
     session = onnxruntime.InferenceSession(onnx_path, providers=providers, provider_options=provider_options)
-    ort_input = {session.get_inputs()[0].name: input_img.cpu().numpy()}
-    ort_output = session.run(None, ort_input)[0]
+    ort_input = {session.get_inputs()[0].name: input_img.cpu().numpy().transpose(0,2,3,1)}
+    ort_output = session.run(None, ort_input)[0].transpose(0,3,1,2)
     if isinstance(ort_output, (tuple, list)):
         ort_output = ort_output[0]
 

test_onnx.py

@@ -22,7 +22,7 @@ class Configs():
         # dataset
 
         parser.add_argument('--dataset', type=str, default='citys', help='dataset name (default: citys)')
-        parser.add_argument('--onnx_path', type=str, default='FPN_int.onnx', help='onnx path')
+        parser.add_argument('--onnx_path', type=str, default='FPN_int_NHWC.onnx', help='onnx path')
         parser.add_argument('--num-classes', type=int, default=19,
                             help='the classes numbers (default: 19 for cityscapes)')
         parser.add_argument('--test-folder', type=str, default='./data/cityscapes',
@@ -78,8 +78,8 @@ def eval_miou(data,path="FPN_int.onnx", device='cpu'):
 
     for i, (image, target) in enumerate(tbar):
         image, target = image.to(device), target.to(device)
-        ort_input = {session.get_inputs()[0].name: image.cpu().numpy()}
-        ort_output = session.run(None, ort_input)[0]
+        ort_input = {session.get_inputs()[0].name: image.cpu().numpy().transpose(0,2,3,1)}
+        ort_output = session.run(None, ort_input)[0].transpose(0,3,1,2)
         if isinstance(ort_output, (tuple, list)):
             ort_output = ort_output[0]
         ort_output = torch.from_numpy(ort_output).to(device)