Evaluation server is live! (and the test set json file is available)

[Ross Girshick]

Greetings!

The evaluation server, hosted on EvalAI, is now open for submissions at https://evalai.cloudcv.org/web/challenges/challenge-page/442/overview. Please see the "Phases" tab for details on submitting results to the Val Phase (to verify that your output format is correct) and the Test Phase (to enter the challenge).

The test set json file that specifies which images are in the test set is now available for download from https://www.lvisdataset.org/dataset.

Best regards,

The LVIS Team (Agrim, Piotr, Ross)

[twa...@gmail.com]

Hi! I'm wondering is there some instructions for the result format?

[twa...@gmail.com]

Hi! I'm wondering is there some instructions for the result format? As I didn't find the description of file formating for uploading in the the "Phases" tab

On Wednesday, September 25, 2019 at 12:23:12 AM UTC+8, Ross Girshick wrote:

[Ross Girshick]

The results format for LVIS is the same as the object detection results format used by COCO, which is documented here: http://cocodataset.org/#format-results. We provide an example of a small results file (on 100 images) in the LVIS API here: https://github.com/lvis-dataset/lvis-api/blob/master/data/lvis_results_100.json.

[张松阳]

Hi, Ross,

Should we submit the  segm results or detection results?

The example in the github is bbox results while this challenge is instance segmentation?

I think it is a little confused.

Thanks?

[Ross Girshick]

Please submit segm (instance segmentation) results. The challenge will be based only on segmentation, not box prediction. Sorry for the confusion from the example results file on github.

Best,

Ross

[莊機動]

Why is the leaderboard not updating? 

Ross Girshick於 2019年10月6日星期日 UTC+8上午7時41分27秒寫道：

[Ross Girshick]

Please see the information in the Phases tab on the EvalAI challenge page. It explains that results are only visible to you (not public).