广东工业智造大赛--赛场一 布匹瑕疵检测

Guangdong Industrial Intelligent Manufacturing Contest--Stage 1

​ 这是我第一次参加天池的大赛，半决赛的代码开源在了final_commit文件夹里面包含了填鸭的代码，第一版的填鸭，计算了patch块的相似度。第二版的我们对小目标（1-4）类的随机3-5倍的放大。 ​ 1.半决赛的Rank：34/100 ​ 2.线下的map：60%左右，线上的map：40%左右

This is the first time I participated in the Tianchi competition. The semi-final code is open sourced. The final_commit folder contains the code for duck filling. In the second version, we zoomed in randomly 3-5 times on the small target (1-4).

1. Rank of semi-final: 34/100

2. Offline Map: about 60%, online map: about 40%

数据(Data)

总结一下(Summary)：

1. 特征工程

2. 选用模型

3. 训练，调参

4. 提交结果

   开始做的时候，我们是先出了一个baseline的结果，开始我是自己一个人玩，直接上faster-rcnn-r101 map：26% 有点沮丧 毕竟这时候还在忙着写（水）论文，后来论坛里 开源了一个cascade-rcnn-r50的模型。初赛52%map。 根据这个baseline，我换了backbone r101 居然：54%map，嘻嘻...这里直接就进了60多。

   Feature engineering

   2. ** Selection model **

   3. ** Training and Tuning **

   4. ** Submission Results **

      **When we started, we first produced a baseline result. At the beginning, I played by myself, and went directly to the faster-rcnn-r101 map: 26% a bit frustrated ** **After all, at this time, I was still busy writing a (water) dissertation. Later, a cascade-rcnn-r50 model was open sourced in the forum. 52% map for the preliminary round. ** According to this baseline, I changed the backbone r101: 54% map, ... more than 60 directly entered here .

有用的点子(effective points)：

• anchor的设计非常重要，mmdetection的默认[0.5,1,2]一般来说很难符合数据的特性，所以这里是提分的点子
• fpn层 dcn （槽点，太吃显存，因为要用很多的offsets）
• OHEM 在线困难样本的发掘
• soft-nms 提分不多，大概一个点左右（大概率是0.几%哈~）
• TTA 老版本的mmdet没有TTA 多尺度测试，新版的有
• 填鸭，对于正常样本的利用。这里其实跟我写论文里的东西有点像，来自小样本增强的那篇论文，但是有个问题就是容易引入结构化的噪声，这里需要计算 patch的块的相似度---于是乎（度娘了一下），用了现成的
• GN+ws （分组Normalize）
• rpn 调参

-The design of the anchor is very important. The default [mm, detection] of mmdetection is generally difficult to meet the characteristics of the data, so here is a point to improve

fpn layer dcn (slot, too much video memory, because a lot of offsets are used)

Discovery of OHEM online difficult samples

Soft-nms does not improve much, about one point (large probability is 0. several% ha ~)

-TTA old version of mmdet does not have TTA multi-scale test, new version

Duck filling, use of normal samples. This is actually a bit like what I wrote in the paper, from the paper with small sample enhancement, but there is a problem that it is easy to introduce structured noise. Here you need to calculate the similarity of the patch blocks-so it ’s (Baidu (google) A little bit), using ready-made

GN + ws (group Normalize)

rpn tuning

一些试过但是没有用的(useless):

• 打算切图做的，cascade-r50对512*512的图像来测试，对于1-4好像有点效果，可能是可视化的错觉吧

• 分类的模型，这时候就做的比较晚，用了 20层的 只有40多的Acc

-I plan to cut the picture. Cascade-r50 tests on 512 * 512 images. It seems to be a bit effective for 1-4 classes. It may be a visual illusion.

-The classification model, which was done relatively late at this time, used 20 layers and only had more than 40 Acc+

还没有尝试的(Haven't tried)：

• GIOU loss

• GMloos GMHRLoss

• Focal loss

• mixup、smooth label

• Cosine annealing learning rate decay

定制化的框架（可能是我没有仔细阅读过mmdet的源码）：

Customized framework (maybe I haven't read the mmdet source code carefully)

• 其实需要魔改框架的，加入一些对小目标增强的结构
• ** In fact, it is necessary to magically change the framework, and add some structures that enhance small goals **

关于训练时多目标（200+）爆显存

About multi-target (200+) explode memory during training

• 参考上传的 transform.py 替换 mmdet的同名文件
• ** Refer to uploaded transform.py to replace mmdet's file with the same name **

Installsion

1. Install

   ​ Fellow the mmdetection install.md

2. Data format

   ​ In my experiment COCO, VOC changed by yourself!

3. COCO pretrained model transfer

   ​ The transfer code in checkpoints num_class should modify your class. Notice!!! scale and ratios changed, I use a simple cat method for suitable model struct param avoid parameters initialize problem!

Training

python3 configs/cascade_rcnn_r101_fpn_1x_with_coco.py  --gpus 1 work_dir XXXXXX(your path to save model and train log)

Test

python3 cascade_rcnn_r101_fpn_1x_test_coco.py

Acknownledgement:

​ Thanks TianChi for holding this competition https://tianchi.aliyun.com/

​ Thanks Openbyes provide computing power support https://openbayes.com/

​ Thanks Team member Chen and Li give valuable advice

如果你有什么问题，欢迎提 issue！

If you have any questions, welcome to issue!

Email：iloveitre@gmail.com