fix backend

Author: Tau-J

projects/rtmpose/rtmdet/person/rtmdet_nano_320-8xb32_coco-person.py

@@ -29,7 +29,7 @@
         max_per_img=100))
 
 train_pipeline = [
-    dict(type='LoadImageFromFile', backend_args={{_base_.backend_args}}),
+    dict(type='LoadImageFromFile'),
     dict(type='LoadAnnotations', with_bbox=True),
     dict(
         type='CachedMosaic',
@@ -53,7 +53,7 @@
 ]
 
 train_pipeline_stage2 = [
-    dict(type='LoadImageFromFile', backend_args={{_base_.backend_args}}),
+    dict(type='LoadImageFromFile'),
     dict(type='LoadAnnotations', with_bbox=True),
     dict(
         type='RandomResize',
@@ -71,7 +71,7 @@
 ]
 
 test_pipeline = [
-    dict(type='LoadImageFromFile', backend_args={{_base_.backend_args}}),
+    dict(type='LoadImageFromFile'),
     dict(type='Resize', scale=(input_shape, input_shape), keep_ratio=True),
     dict(
         type='Pad',