在tensorflow上跑yolo

1. 下载源码
   
   git clone https://github.com/hizhangp/yolo_tensorflow

   1 2	git clone https://github.com/hizhangp/yolo_tensorflow

2. 创建数据集和权重目录
   
   cd yolo_tensorflow mkdir data cd data mkdir pascal_voc mkdir weights

   1 2 3 4 5 6

   cd yolo_tensorflow mkdir data cd data mkdir pascal_voc mkdir weights

3. 下载数据集
   
   wget http://pjreddie.com/media/files/VOCtrainval_06-Nov-2007.tar tar xvf VOCtrainval_06-Nov-2007.tar mv VOCdevkit pascal_voc/.

   1 2 3 4

   wget http://pjreddie.com/media/files/VOCtrainval_06-Nov-2007.tar tar xvf VOCtrainval_06-Nov-2007.tar mv VOCdevkit pascal_voc/.

4. 下载权重
   
   tar zxvf YOLO_small.tar.gz mv YOLO_small.ckpt yolo_tensorflow/data/weights

   1 2 3

   tar zxvf YOLO_small.tar.gz mv YOLO_small.ckpt yolo_tensorflow/data/weights

5. 测试
   
   cd yolo_tensorflow python test.py

   1 2 3

   cd yolo_tensorflow python test.py

   
   

   

出现的问题

1. 报错AttributeError: module ‘tensorflow.python.ops.nn’ has no attribute ‘leaky_relu’solution：修改yolo/yolo_net 244行
   
   return tf.nn.leaky_relu(inputs, alpha=alpha, name='leaky_relu')

   1 2	return tf.nn.leaky_relu(inputs, alpha=alpha, name='leaky_relu')

   
   换成
   
   return tf.maximum(inputs, alpha * inputs)

   1 2	return tf.maximum(inputs, alpha * inputs)