fix md5 of fruit dataset (#3863)

PaddleCV/PaddleDetection/docs/QUICK_STARTED.md

@@ -26,7 +26,7 @@ Training:
 python -u tools/train.py -c configs/yolov3_mobilenet_v1_fruit.yml \
                         --use_tb=True \
                         --tb_log_dir=tb_fruit_dir/scalar \
-                        --eval \
+                        --eval
 ```
 
 Use `yolov3_mobilenet_v1` to fine-tune the model from COCO dataset. Meanwhile, loss and mAP can be observed on tensorboard.  

PaddleCV/PaddleDetection/docs/QUICK_STARTED_cn.md

@@ -26,7 +26,7 @@ export CUDA_VISIBLE_DEVICES=0
 python -u tools/train.py -c configs/yolov3_mobilenet_v1_fruit.yml \
                         --use_tb=True \
                         --tb_log_dir=tb_fruit_dir/scalar \
-                        --eval \
+                        --eval
 ```
 
 训练使用`yolov3_mobilenet_v1`基于COCO数据集训练好的模型进行finetune。训练期间可以通过tensorboard实时观察loss和精度值，启动命令如下：

PaddleCV/PaddleDetection/ppdet/utils/download.py

@@ -74,6 +74,7 @@ DATASETS = {
     'fruit': ([(
         'https://dataset.bj.bcebos.com/PaddleDetection_demo/fruit-detection.tar',
         'ee4a1bf2e321b75b0850cc6e063f79d7', ), ], ["fruit-detection"]),
+    'objects365': (),
 }
 
 DOWNLOAD_RETRY_LIMIT = 3
@@ -104,6 +105,11 @@ def get_dataset_path(path, annotation, image_dir):
         if os.path.split(path.strip().lower())[-1] == name:
             logger.info("Parse dataset_dir {} as dataset "
                         "{}".format(path, name))
+            if name == 'objects365':
+                raise NotImplementedError(
+                    "Dataset {} is not valid for download automatically."
+                    "Please apply and download the dataset from."
+                    "https://www.objects365.org/download.html")
             data_dir = osp.join(DATASET_HOME, name)
 
             # For voc, only check dir VOCdevkit/VOC2012, VOCdevkit/VOC2007