update pretrained models & fix pycocotools

1. add pretrained resnet152, snetv2_0.5x, 1.5x and 2.0x
2. move `import pycocotools` inside `main` to prevernt import error

README.md

@@ -60,8 +60,12 @@ export PYTHONPATH=/path/to/models:$PYTHONPATH
 | ResNet34 |  73.960  |  91.630  |
 | ResNet50 | 76.254 | 93.056 |
 | ResNet101 | 77.944 | 93.844 |
+| ResNet152 | 78.582 | 94.130 | 
 | ResNeXt50 32x4d | 77.592 | 93.644 |
-| ShuffleNetV2 x1.0 |  69.369  |  88.793  |
+| ShuffleNetV2 x0.5 |  60.696  |  82.190  | 
+| ShuffleNetV2 x1.0 |  69.372  |  88.764  | 
+| ShuffleNetV2 x1.5 |  72.806  |  90.792  | 
+| ShuffleNetV2 x2.0 |  75.074  |  92.278  | 
 
 
 ### 目标检测

official/vision/classification/resnet/README.md

@@ -12,6 +12,7 @@
 | ResNet34 |  73.960  |  91.630  | 
 | ResNet50 | 76.254 | 93.056 | 
 | ResNet101 | 77.944 | 93.844 | 
+| ResNet152 | 78.582 | 94.130 | 
 | ResNeXt50 32x4d | 77.592 | 93.644 | 
 
 用户可以通过`megengine.hub`直接加载本目录下定义好的模型，例如：

official/vision/classification/resnet/model.py

@@ -341,6 +341,9 @@ def resnet101(**kwargs):
     return ResNet(Bottleneck, [3, 4, 23, 3], **kwargs)
 
 
+@hub.pretrained(
+    "https://data.megengine.org.cn/models/weights/resnet152_fbaug_78582_7551aff3.pkl"
+)
 def resnet152(**kwargs):
     r"""ResNet-152 model from
     `"Deep Residual Learning for Image Recognition" <https://arxiv.org/pdf/1512.03385.pdf>`_

official/vision/classification/shufflenet/README.md

@@ -8,7 +8,10 @@
 
 | 模型 | top1 acc | top5 acc |
 | --- | --- | --- |
-| shufflenet_v2_x1_0 |  69.369  |  88.793  | 
+| ShuffleNetV2 x0.5 |  60.696  |  82.190  | 
+| ShuffleNetV2 x1.0 |  69.372  |  88.764  | 
+| ShuffleNetV2 x1.5 |  72.806  |  90.792  | 
+| ShuffleNetV2 x2.0 |  75.074  |  92.278  | 
 
 用户可以通过`megengine.hub`直接加载本目录下定义好的模型，例如：
 

official/vision/classification/shufflenet/model.py

@@ -216,10 +216,16 @@ class ShuffleNetV2(M.Module):
                     M.init.fill_(m.bias, 0)
 
 
+@hub.pretrained(
+    "https://data.megengine.org.cn/models/weights/snetv2_x2_0_75115_497d4601.pkl"
+)
 def shufflenet_v2_x2_0(num_classes=1000):
     return ShuffleNetV2(num_classes=num_classes, model_size="2.0x")
 
 
+@hub.pretrained(
+    "https://data.megengine.org.cn/models/weights/snetv2_x1_5_72775_38ac4273.pkl"
+)
 def shufflenet_v2_x1_5(num_classes=1000):
     return ShuffleNetV2(num_classes=num_classes, model_size="1.5x")
 
@@ -231,5 +237,8 @@ def shufflenet_v2_x1_0(num_classes=1000):
     return ShuffleNetV2(num_classes=num_classes, model_size="1.0x")
 
 
+@hub.pretrained(
+    "https://data.megengine.org.cn/models/weights/snetv2_x0_5_60750_c28db1a2.pkl"
+)
 def shufflenet_v2_x0_5(num_classes=1000):
     return ShuffleNetV2(num_classes=num_classes, model_size="0.5x")

official/vision/detection/tools/test.py

@@ -21,8 +21,6 @@ from megengine import jit
 from megengine.data import DataLoader, SequentialSampler
 from megengine.data.dataset import COCO as COCODataset
 from tqdm import tqdm
-from pycocotools.coco import COCO
-from pycocotools.cocoeval import COCOeval
 
 from official.vision.detection.tools.nms import py_cpu_nms
 
@@ -272,6 +270,9 @@ def make_parser():
 
 
 def main():
+    from pycocotools.coco import COCO
+    from pycocotools.cocoeval import COCOeval
+    
     parser = make_parser()
     args = parser.parse_args()