VisDrone Part

Try to add our email address to your email whitelist: aiskyeye@qq.com and notification@aiskyeye.com. Alternatively, you can leave a message below to request manual account activation.
The bounding boxes in the submitted results overlapping the “others” objects with the threshold larger than the threshold or locating in the ignored regions will be filtered out in the evaluation.
NO. If you would like to report results of your algorithm with different settings (e.g., different parameters and training conditions), please use the training and validation sets for this purpose, and only submit one result to the evaluation server.
Similar to the PASCAL VOC dataset, we labeled some “ignore” regions, which indicates the regions containing the objects that are difficult to be annotated due to low resolution or crowd. Meanwhile, some rarely occurring objects (e.g., machineshop truck, forklift truck, and tanker) are labeled as “others”, which are ignored in evaluation.
Thank you! We are aware of some deficiencies in the provided annotations and are still trying to improve these. We do appreciate all kinds of feedback, so please don’t hesitate to contact us to report any findings.
The evaluation results will be displayed at the bottom of the submission page, and the leaderboard will be updated daily.
I you want to rename your algorithm, please contact us.

185 Comments

  1. Sorry, I have two more quesions.
    In the SUBMIT page , when I choose the ” OBHECT DETECTION” it shows”Your upload quota has been used up.”
    But actually I have just downloaded the test dataset, I have never submitted any file.
    When I choose the “Single-Object Tracking”, it comes to my accout page. Its really wired. Could u please help me?

    My second question is , could you please tell me the files in “VisDrone2019-SOT-test-dev/attributes/*attr.txt” What do they mean,especially the numbers like”0,0,1,0,0,0,1,0,0,0,1,0″. What do these parameters mean respectively. I didn’t find any document to give a definition. thank you

  2. I have two questions.
    First, the result format of the crowd counting task is the format of object detection, not the format of mentioned above. Could you tell me the proper format?
    Second, the EVALUATE directory tells me “The evaluation code for crowd counting is available on the VisDrone github”, but in this github I can not find eavluation code of crowd counting. Could you please help me ?

  3. Hi, I have three questions:

    1. If I understand correctly, each company/group has 3 test submissions per algorithm. So if there would be totally different algorithms used (not just some hyperparameters that change), one company/group would have 3 submissions per algorithm?

    2. Is it 3 (like in the FAQ) or 5 (like on the submission page) submissions per algorithm?

    3. How is the test-dev evaluation working? Just locally on my machine or is there a seperate evaluation page as well?
    Thanks in advance!

    1. Hello, we have updated the old FAQ page.

      1. In the ECCV2020 challenge, each company/group has 5 test submissions totally. We recommend that each team only participate in one task. If your team wants to participate in multiple tasks, it is recommended to register a new account to get more submission opportunities, or contact us to modify your account upload limit.

      2. The upload quota on the submission page is all your submission quota (shared in all tasks)

      3. Evaluation is done on our server. You need to download the test set, get the results locally, and submit the test set results to our server. The evaluation will be performed automatically. After the evaluation is completed, you will receive a notification email and the score will be displayed below the submission form.

  4. I have a question.
    The description document of the corresponding method is asked to submit when I submit the results.But in your download link, I don’t find document templates. There is only one download link for the dataset.
    Could you help me?

  5. Hi, “Please note that when you submit the results, you must also submit the corresponding method description document. Document templates can be downloaded on the download page. ” in Submit Page
    The download page is not found!

  6. How to evaluate test-dev result on evaluation server? In webpage you guys said “Number of time for testing on evaluation server for test-dev dataset is unlimited” so it is allowed to test test-dev set right? But I did not see any info. Please let me know how to do it.

  7. Hi,
    I haven’t seen any discrimination for short-term and long-term trackers. Please let me know if we are allowed to use a detector/verifier for our short-term tracker in the VisDrone2020 challenge.

    Mojtaba
  8. Hello, I am a member participating in 2020SOT. I submitted the zip in the format you gave, but the total submission was unsuccessful. Trouble to ask why. See the attachment for a schematic diagram of the submission format. Thank you.

  9. I can see having validation and test-dev datasets excluded from the training process for debugging/offline evaluation purposes. Regarding the final submission on the test-challenge set, are we supposed to train on the whole dataset including all the available annotations (train, val and test-dev) or we should rely on the trainining-set split explicitly?

    Ioannis
  10. Hello, I have a question about the paper submitition. If I want to submit a paper and have good results on serveral standard datasets, do I have to join your challenge and have results on visdrone dataset, or I just need to fit in one of your topics, for instance, people and object tracking?

    yukawa
  11. There are many problems:
    1. Whether the pdf version of the description cannot be included in the submitted documents, otherwise there will be errors in the documents;
    2. Why does the provided counting document template include 00011.tex file

  12. In the register rules, we have:

    “`
    1. Please register by the team leader first, choose Role as the team leader, and then let your teammates register, fill in the Team leader Username, and only the team leader has the qualifications for submission.
    “`

    – It means only leader can submit the result?
    – Can member submit other challenge which is different from leader’s challenge?

  13. Hi,
    I have created the submission file for the crowd counting task as per the guidelines but the server message says format error. I have tried submitting multiple times still it shows the same error. I have mailed you my result file, please look into it.

  14. Hi, I have encountered some problems with the results I submitted on the MOT. For example, when viewing locally, my tracker visualization is very good, but the evaluation indicators after submitting to the platform are very low. I checked the format that needs to be submitted and found no problems. I really want to ask what the evaluation mechanism of the platform is, the target_id needs to be set from 0, or whether there will be some evaluation errors on the platform. If you can see these, I very much hope to give me some replies and suggestions.

    cjy
  15. Hi,
    While submitted the results, it displayed “uploaded successfully” but under YOUR SUBMISSIONS section it shows “upload” under the status bar and “N/A” under the score bar. How should i know the exact status of submission and the scores? Should i wait or re-submit?

  16. hello,
    i want to use your dataset, VisDrone2019, as a benchmark to eval my objection detector, but i cannot find an complete intro to your annotations of the dataset.i know the the 1st to the 4th items are the localization information of a bbox, the 6th is the category, and the last two are the occlusion ratio and truncation ratio, so what’s the 5th item?

    Dan
    1. Hi,
      The account itself belongs to the team leader, and the name filled in at registration is the name of the team leader.
      The information filled in the “Team members” column is the information of the team members, and there is no need to confirm whether they have joined.

    1. 您好!我们使用的网站邮件服务给了我们这个提示,也许您需要换个邮箱注册,抱歉。

      无法发送到 liukewei@hud.edu.cn
      退信原因 收件人(liukewei@hud.edu.cn)所属域名不存在,邮件无法送达。
      No Mx Record Found
      解决方案 请联系您的收件人,重新核实邮箱地址,或发送到其他收信邮箱。 您也可以向管理员报告此退信。

  17. Hello, organizers:
    I have now registered an account, so how do my other team members need to join my squad? I don’t seem to find a similar option after I log in. What do my teammates need to do when they register?
    I didn’t find any clear information on the official website, can you please answer this? Thanks!

    1. 参赛者,您好!
      已经注册的队员不使用评测次数就不会违反规则。
      队员名单在最终成绩公布后会有一次核实机会,请不用担心。

      Hello, participants!
      Registered players do not violate the rules if they do not use the chances of evaluation.
      There will be a chance to verify team members after the final results are announced, so please don’t worry.

  18. 您好,关于任务 1我有几点想明确的地方:
    1、任务概述中说是对十个预定义类的检测,提交格式页面类别字段有0-11类,其中0是忽略区域,11是稀有类,在算法中我们是不是只需要检测类1-10即可
    2、每一张图片的检测结果是否是按置信度排序从高到低写入txt

  19. 请问测试每个图片的测试结果必须为500个框吗 在用官方给的evalDet.m测试test-dev的指标(其中很多预测结果只保留了比较好的预测结果,所以目标数小于500),但是指标测试程序报错了,所以test-challenge上是要每个图片要有500行结果?

回复 匿名 取消回复

您的电子邮箱地址不会被公开。