search

zzzxxxttt / pytorch_simple_CenterNet_45

A simple pytorch implementation of CenterNet (Objects as Points)
308 stars 64 forks source link

Result Extraction #28

Open JiaLim98 opened 4 years ago

JiaLim98 commented 4 years ago

Hi @zzzxxxttt,

Thank you for your amazing work. I have completed my training and testing. However, I require two more results for documentation purposes.

  1. Precision-Recall Curve: May I know which python file is the repo using? This is because I found two cocoeval.py after finish compiling PythonAPI, under the two paths stated below. I have tried modifying both but can't seem to work (looks like the test.py is not using both) . Can you let me know how can I extract the 101 datapoints for PR curve?
    • \lib\cocoapi\PythonAPI\build\lib.linux-x86_64-3.6\pycocotools, and
    • \lib\cocoapi\PythonAPI\pycocotools
  2. mAP@0.5 by classes. I can't seem to find the line of code that gives me this. Can you let me know how? I got the overall coco results already, I just need the results by classes.

Hope to hear from you soon. Thank you.

Cheers, JiaLim98

zzzxxxttt commented 4 years ago

For the pr curve, this issue may help, and detectron2 generates class-wise mAP during evaluation, you can check the implementation of it.

JiaLim98 commented 4 years ago

Thanks for the reply.

For anyone who wants to output class-wise mAP, this is the most straightforward way I found. The mAP for entire model is calculated from the average of APs for each classes. I have found the code for exporting the class-wise mAP from cocoeval.py in Pycocotools here from line 458 to line 464.

yangyang117 commented 3 years ago

Jialim，I am very happy that you have completed the training and testing work, may I know your final accuracy? I am trying to train this net in my datasets, but I still not complete it.

JiaLim98 commented 3 years ago

Hi @yangyang117,

My final accuracy (without finetuning) are: mAP@0.5 = 92.02% and mAP@0.5:0.05:0.95 = 40.60% Hope it helps.

yangyang117 commented 3 years ago

Thanks for your answering, is this for your own datasets or public datasets?

JiaLim98 commented 3 years ago

It is an open-sourced PCB defects dataset.

yangyang117 commented 3 years ago

Thank you, it's very helpful.

18342960922 commented 3 years ago

它是开源的PCB缺陷数据集。

您好，麻烦问一下 自己的数据集 遇到这个问题 No such file or directory: './data/voc/VOCdevkit/VOC2007/annotations_cache'该如何处理

JiaLim98 commented 3 years ago

您可以检查文件目录是否确实存在. 对于第一次训练，这个问题不应该存在.