专栏首页深度学习计算机视觉无人机图片物体检测baseline

无人机图片物体检测baseline

<object_category>

The object category indicates the type of annotated object, (i.e., ignored regions (0), pedestrian (1), people (2), bicycle (3), car (4), van (5), truck (6), tricycle (7), awning-tricycle (8), bus (9), motor (10), others (11))

Backbone = Resnext 152

test 548, step =50000, mask_on=False 测548) INFO voc_dataset_evaluator.py: 143: Mean AP = 0.4277 INFO voc_dataset_evaluator.py: 144: ~~~~~~~~ INFO voc_dataset_evaluator.py: 145: Results: INFO voc_dataset_evaluator.py: 147: 0.407 INFO voc_dataset_evaluator.py: 147: 0.345 INFO voc_dataset_evaluator.py: 147: 0.234 INFO voc_dataset_evaluator.py: 147: 0.771 INFO voc_dataset_evaluator.py: 147: 0.448 INFO voc_dataset_evaluator.py: 147: 0.419 INFO voc_dataset_evaluator.py: 147: 0.397 INFO voc_dataset_evaluator.py: 147: 0.182 INFO voc_dataset_evaluator.py: 147: 0.568 INFO voc_dataset_evaluator.py: 147: 0.506 INFO voc_dataset_evaluator.py: 149: ~~~~~~

+mask on (40000,测25, lr=0.01,WEIGHT_DECAY=0.00001) INFO voc_dataset_evaluator.py: 144: AP for pedestrian = 0.4490 INFO voc_dataset_evaluator.py: 144: AP for people = 0.3275 INFO voc_dataset_evaluator.py: 144: AP for bicycle = 0.2681 INFO voc_dataset_evaluator.py: 144: AP for car = 0.7616 INFO voc_dataset_evaluator.py: 144: AP for van = 0.4760 INFO voc_dataset_evaluator.py: 144: AP for truck = 0.4568 INFO voc_dataset_evaluator.py: 144: AP for tricycle = 0.5157 INFO voc_dataset_evaluator.py: 144: AP for awning-tricycle = 0.1051 INFO voc_dataset_evaluator.py: 144: AP for bus = 0.6198 INFO voc_dataset_evaluator.py: 144: AP for motor = 0.4912 INFO voc_dataset_evaluator.py: 147: Mean AP = 0.4471

+mask on (30000,测25, lr=0.01,WEIGHT_DECAY=0.00001) INFO voc_dataset_evaluator.py: 144: AP for pedestrian = 0.4526 INFO voc_dataset_evaluator.py: 144: AP for people = 0.3127 INFO voc_dataset_evaluator.py: 144: AP for bicycle = 0.1926 INFO voc_dataset_evaluator.py: 144: AP for car = 0.7616 INFO voc_dataset_evaluator.py: 144: AP for van = 0.4662 INFO voc_dataset_evaluator.py: 144: AP for truck = 0.4802 INFO voc_dataset_evaluator.py: 144: AP for tricycle = 0.4538 INFO voc_dataset_evaluator.py: 144: AP for awning-tricycle = 0.2240 INFO voc_dataset_evaluator.py: 144: AP for bus = 0.6970 INFO voc_dataset_evaluator.py: 144: AP for motor = 0.4490 INFO voc_dataset_evaluator.py: 147: Mean AP = 0.4490

+mask on (20000,测25, lr=0.01,WEIGHT_DECAY=0.00001) INFO voc_dataset_evaluator.py: 144: AP for pedestrian = 0.4629 INFO voc_dataset_evaluator.py: 144: AP for people = 0.3280 INFO voc_dataset_evaluator.py: 144: AP for bicycle = 0.2711 INFO voc_dataset_evaluator.py: 144: AP for car = 0.7682 INFO voc_dataset_evaluator.py: 144: AP for van = 0.5189 INFO voc_dataset_evaluator.py: 144: AP for truck = 0.4190 INFO voc_dataset_evaluator.py: 144: AP for tricycle = 0.4645 INFO voc_dataset_evaluator.py: 144: AP for awning-tricycle = 0.2197 INFO voc_dataset_evaluator.py: 144: AP for bus = 0.6752 INFO voc_dataset_evaluator.py: 144: AP for motor = 0.5106 INFO voc_dataset_evaluator.py: 147: Mean AP = 0.4638

+mask on (20000,测548, lr=0.01,WEIGHT_DECAY=0.00001) INFO voc_dataset_evaluator.py: 144: AP for pedestrian = 0.4756 INFO voc_dataset_evaluator.py: 144: AP for people = 0.3603 INFO voc_dataset_evaluator.py: 144: AP for bicycle = 0.2444 INFO voc_dataset_evaluator.py: 144: AP for car = 0.7805 INFO voc_dataset_evaluator.py: 144: AP for van = 0.4889 INFO voc_dataset_evaluator.py: 144: AP for truck = 0.4200 INFO voc_dataset_evaluator.py: 144: AP for tricycle = 0.3882 INFO voc_dataset_evaluator.py: 144: AP for awning-tricycle = 0.2106 INFO voc_dataset_evaluator.py: 144: AP for bus = 0.6713 INFO voc_dataset_evaluator.py: 144: AP for motor = 0.4979 INFO voc_dataset_evaluator.py: 147: Mean AP = 0.4538

+mask on (50000,测548, lr=0.01,WEIGHT_DECAY=0.00001) INFO voc_dataset_evaluator.py: 144: AP for pedestrian = 0.4511 INFO voc_dataset_evaluator.py: 144: AP for people = 0.3418 INFO voc_dataset_evaluator.py: 144: AP for bicycle = 0.2552 INFO voc_dataset_evaluator.py: 144: AP for car = 0.7776 INFO voc_dataset_evaluator.py: 144: AP for van = 0.4749 INFO voc_dataset_evaluator.py: 144: AP for truck = 0.4285 INFO voc_dataset_evaluator.py: 144: AP for tricycle = 0.3880 INFO voc_dataset_evaluator.py: 144: AP for awning-tricycle = 0.1663 INFO voc_dataset_evaluator.py: 144: AP for bus = 0.6388 INFO voc_dataset_evaluator.py: 144: AP for motor = 0.4513 INFO voc_dataset_evaluator.py: 147: Mean AP = 0.4373

+mask on (30000,测548, lr=0.01,WEIGHT_DECAY=0.00001) INFO voc_dataset_evaluator.py: 144: AP for pedestrian = 0.4586 INFO voc_dataset_evaluator.py: 144: AP for people = 0.3270 INFO voc_dataset_evaluator.py: 144: AP for bicycle = 0.2106 INFO voc_dataset_evaluator.py: 144: AP for car = 0.7818 INFO voc_dataset_evaluator.py: 144: AP for van = 0.4778 INFO voc_dataset_evaluator.py: 144: AP for truck = 0.4534 INFO voc_dataset_evaluator.py: 144: AP for tricycle = 0.3711 INFO voc_dataset_evaluator.py: 144: AP for awning-tricycle = 0.1745 INFO voc_dataset_evaluator.py: 144: AP for bus = 0.6454 INFO voc_dataset_evaluator.py: 144: AP for motor = 0.4528 INFO voc_dataset_evaluator.py: 147: Mean AP = 0.4353

+mask on (40000,测548, lr=0.01,WEIGHT_DECAY=0.00001) INFO voc_dataset_evaluator.py: 144: AP for pedestrian = 0.4680 INFO voc_dataset_evaluator.py: 144: AP for people = 0.3339 INFO voc_dataset_evaluator.py: 144: AP for bicycle = 0.2140 INFO voc_dataset_evaluator.py: 144: AP for car = 0.7810 INFO voc_dataset_evaluator.py: 144: AP for van = 0.4537 INFO voc_dataset_evaluator.py: 144: AP for truck = 0.4579 INFO voc_dataset_evaluator.py: 144: AP for tricycle = 0.3984 INFO voc_dataset_evaluator.py: 144: AP for awning-tricycle = 0.1446 INFO voc_dataset_evaluator.py: 144: AP for bus = 0.6712 INFO voc_dataset_evaluator.py: 144: AP for motor = 0.5066 INFO voc_dataset_evaluator.py: 147: Mean AP = 0.4429

Backbone = Resnet 101

Backbone = Resnext 101

本文参与腾讯云自媒体分享计划,欢迎正在阅读的你也加入,一起分享。

我来说两句

0 条评论
登录 后参与评论

相关文章

  • 白盒测试中的几种覆盖方法

    文章转自:http://www.51testing.com/html/44/n-3713444.html 白盒测试用例设计的一个很重要的评估标准就是对代码的覆...

    张俊怡
  • synchronized与lock

    synchronized是java中的一个关键字,也就是说是Java语言内置的特性,而Lock是后来出现的类 Lock和synchronized相比有更多的功...

    张俊怡
  • 何恺明等NeurlPS新作:定义迁移学习新范式

    【摘要】基于深度学习的迁移学习的主流方法一般是从一个任务中学习到可迁移到其他任务的通用特征向量,例如语言中的单词嵌入和视觉中的预训练卷积特征(比如imagene...

    张俊怡
  • bootstrap-table表格超过长度自动隐藏

    今天说一说bootstrap-table表格优化相关的,当表格里面的内容列数过多的,内容长度过长的时候,会出现自动换行的效果,整个表格的每一行高度就不相同了,看...

    祈澈菇凉
  • ASP.NET MVC 5 - 将数据从控制器传递给视图

    在我们讨论数据库和数据模型之前,让我们先讨论一下如何将数据从控制器传递给视图。控制器类将响应请求来的URL。控制器类是给您写代码来处理传入请求的地方,并从数据库...

    葡萄城控件
  • 分割中英文,并且中文不能分割一半?

    题目:编写一个截取字符串的函数,输入为一个字符串和字节数,输出为按字节截取的字符串。但是要保证汉字不被截半个,如“我ABC”4,应该截为“我AB”,输入“我AB...

    格姗知识圈
  • 如何管理Spark Streaming消费Kafka的偏移量(二)

    我是攻城师
  • App Store把app的评论扒下来

    链接:https://github.com/freesan44/AppReviews

    freesan44
  • 多线程使用的关键字

      Synchronized修饰一个方法很简单,就是在方法的前面加synchronized,synchronized修饰方法和修饰一个代码块类似,只是作用范围不...

    小勇DW3
  • 让树莓派开机运行Python脚本

    这个方式不用修改 rc.local 文件。机制上类似于 Windows 的“开始”菜单中的“启动”菜单。方法如下:

    十四君

扫码关注云+社区

领取腾讯云代金券

玩转腾讯云 有奖征文活动