yolov3训练误差可视化

通过把yolov3训练出来的效果可视化：

可以用python可视化，也可以直接在代码里面用tensorboardX可视化。

  

   

    

     

    
    

     

      import re
     
    
   

    

     

    
    

     

      import pandas as pd
     
    
   

    

     

    
    

     

      from matplotlib import pyplot as plt
     
    
   

    

     

    
    

     
 
     
    
   

    

     

    
    

     
 
     
    
   

    

     

    
    

     

      def read_data(filename):
     
    
   

    

     

    
    

     
    # 1: 3983.445801, 3983.445801 avg loss, 0.000000 rate, 8.518392 seconds, 128 images
     
    
   

    

     

    
    

     

          pattern = re.compile("^\\s*(\\d+):.*?, (.*?) avg loss")
     
    
   

    

     

    
    

     

          lines = None
     
    
   

    

     

    
    

     
    with open(filename) as r:
     
    
   

    

     

    
    

     

              lines = r.readlines()
     
    
   

    

     

    
    

     
 
     
    
   

    

     

    
    

     

          data = [pattern.findall(line)[0] for line in lines]
     
    
   

    

     

    
    

     

          data = [(int(row[0]), float(row[1])) for row in data]
     
    
   

    

     

    
    

     
    return pd.DataFrame(data, columns=['epoch', 'loss'])
     
    
   

    

     

    
    

     
 
     
    
   

    

     

    
    

     
 
     
    
   

    

     

    
    

     

      def show(filename, n=1):
     
    
   

    

     

    
    

     
    """
     
    
   

    

     

    
    

     

       可视化yolov3训练loss
     
    
   

    

     

    
    

     

       :param filename 文件名
     
    
   

    

     

    
    

     

       :param n 设置若干个epoch的loss合并到一起，在相同区间内取最小值
     
    
   

    

     

    
    

     

       """
     
    
   

    

     

    
    

     

          df = read_data(filename)
     
    
   

    

     

    
    

     

          loss_min = min(df.loss)
     
    
   

    

     

    
    

     
    print("Loss最小值：", loss_min)
     
    
   

    

     

    
    

     
    print("对应epoch：", list(df.epoch[df.loss == loss_min]))
     
    
   

    

     

    
    

     

          epochs = max(df.epoch)
     
    
   

    

     

    
    

     

          y = [min(df.loss[(i*n < df.epoch) & (df.epoch <= (i+1)*n)])
     
    
   

    

     

    
    

     
         for i in range(epochs//n)]
     
    
   

    

     

    
    

     

          plt.title("Yolov3 train loss")
     
    
   

    

     

    
    

     

          plt.xlabel("Epoch")
     
    
   

    

     

    
    

     

          plt.ylabel("loss")
     
    
   

    

     

    
    

     

          plt.plot([(i+1)*n for i in range(epochs//n)], y)
     
    
   

    

     

    
    

     

          plt.show()
     
    
   

    

     

    
    

     
 
     
    
   

    

     

    
    

     
 
     
    
   

    

     

    
    

     

      if __name__ == '__main__':
     
    
   

    

     

    
    

     
    import fire
     
    
   

    

     

    
    

     

          fire.Fire(show)
     
    
  
 

参考链接：https://github.com/IBBD/IBBD.github.io/blob/b4848c548d1e9d56eb213e83f78da0d396c6b89e/machine-learning/yolov3-train-model.md

文章来源: blog.csdn.net，作者：小小谢先生，版权归原作者所有，如需转载，请联系作者。

原文链接：blog.csdn.net/xiewenrui1996/article/details/109278444