Microsoft COCO Image Captioning Challenge Forum

Go back to competition Back to thread list Post in this thread

> Evaluatin Error

The first time I submitted results for test2015, I got the error
Traceback (most recent call last):
File "/codalabtemp/tmpwWHgCE/run/program/cocoEvalCapCoda.py", line 48, in
elif not file.find(valset) == -1 and fileSplit[3] == 'results.json':
IndexError: list index out of range

The second time I got the error
Traceback (most recent call last):
File "/codalabtemp/tmpYeqj3O/run/program/cocoEvalCapCoda.py", line 59, in <module>
res = json.load(open(resFile))
IOError: [Errno 2] No such file or directory: '/codalabtemp/tmpYeqj3O/run/input/res/captions_test2014__results.json'

I did generate captions for all the 81434 test images, the val images are actually also covered by the test image set, what could be the problem?

Posted by: wenhuchen @ Sept. 28, 2016, 9:32 a.m.

Hi, I also got this error. Have you found the solutions?

Posted by: sophieag @ Nov. 21, 2016, 2:50 p.m.

May be you should not contain the folder named "results", just zip the two json files, and named it "results.zip".

Posted by: chenxp2311 @ Jan. 25, 2017, 5:38 a.m.

which two files? i thought only test result file was enough. Pls help!!

Posted by: Rishi @ Jan. 17, 2021, 10:14 a.m.

I got the same error, and the solution posted by chenxp2311 worked for me! Specifically, rather than putting the prediction jsons in a folder called results and running `zip results.zip results`, it works to run `zip results.zip <val_preds> <test_preds>`. Thanks, chenxp2311!

Posted by: amitakamath @ March 2, 2021, 11:27 p.m.

I got the same error. Is there any requirement for the name of <val_preds> <test_preds> .

Posted by: NA @ March 4, 2022, 2:17 p.m.

I solved this error by the following procedure:

1) The <val_preds> and <test_preds> are named as <captions_val2014_method_results.json> and <captions_test2014_method_results.json>.

2) Run `zip results.zip <val_preds> <test_preds>` to get the results.zip.

Posted by: Runbor_Wang @ Dec. 17, 2023, 12:35 p.m.
Post in this thread