Soda Bottle Identification Challenge


#107

Hmm, so the users of desktop software has lower chance to win?
Will you release an update with the fix for thes dekstop users before the competition ends?


#108

I don’t think so. That is a bug in the software, how would that effect evaluation on their test data


#109

@rajendra Sir, please see what’s going on. Why it’s getting so late in evaluation


#110

We have made some changes and submissions will be processed much faster now. One of your submission is already processed.


#111

Yeah just now I received mail sir. Thank you


#112

The bug is only when pretrained model is used, so if you aren’t using it, you are still in the race with equal chances to win :sparkles:

And about the issue, we will soon fix this bug and keep you posted.
Thanks!


#113

Hi! @Architha, I need a clarification regarding the prize. Will the prize be given in form of credits on cloud or in form of money?


#114

@Architha, Yea even i had the same question as @swap_neil


#115

A have switched to cloud and it is a wonderful experince :slight_smile: (app is instantly usable, easy to switch GPU instance ON/OFF, no fan noise ;), and if somebody already tried some pretrained network on the desktop the free 2h can be enough)

I found some useful tips in this video about network training (5th place Kaggle competition): https://www.youtube.com/watch?v=Mw2vdYv4ups&feature=youtu.be


#117

Prizes will be money. Please check all the rules as we will be distributing cash prizes based on the rules mentioned

Thanks
Mahesh


#118

It says package is already installed


#119

@krishnaadithya You can change your code to import cv2 directly (instead of from . import cv2 , you can do import cv2)


#120

@horvathj, @Architha

While we are working on releasing the next version with the fix, here is the patch mentioned in this thread to fix the issue temporarily.


#121

I did import cv2

from . import cv2 is some back ground code
that is the error I get

and is there now way for us to use the keras 2 codes?


#122

Hey deepgcntion team,

just a quick hint.

Currently it is possible do start submitten a file even though I am not logged in. After uploading I get the message that I need to be logged in to sumbit( green colored, as if I uploaded the file sucessfully).

After loggin in, I am back on the deepcognition main page. After this process its not clear to me if I submitted my model or not.

Would be nice, to get an error message bevor the uploading start, quoting I need to log in first.


#123

Thanks for the feedback. We will improve it before the next competition.


#124

why am i always getting 0.125 score despite the model accuracy above 0.95?


#125

check this thread:


#126

do i need to run the same model once more after these changes before submission ?


#127

yes, you need to train the model again after making the changes.