Keras-RetinaNet-for-Open-Images-Challenge-2018
Keras-RetinaNet-for-Open-Images-Challenge-2018 copied to clipboard
ZFTurbo
Hi,I have a error in run inference!
I get this error like this: FileNotFoundError: [Errno 2] No such file or directory: '/home/a/Bcw_data/open-image-challenge/input/class-descriptions-boxable.csv' I want to know the '.csv' is where.
You can download it from here: https://storage.googleapis.com/openimages/v5/class-descriptions-boxable.csv
You can download it from here: https://storage.googleapis.com/openimages/v5/class-descriptions-boxable.csv
Thank you very much!I want use this implement in the 2019 open images challenge.I want to try the result!Thank you for your help!
Warning is ok. You have zero files in folder. Check if there are some JPGs available.
Warning is ok. You have zero files in folder. Check if there are some JPGs available.
yes,I find a error in my test file!I have solved it!Thanks you very much!
Warning is ok. You have zero files in folder. Check if there are some JPGs available.
Hi,I use this implement to test the open Image 2019,I use the resnet101 pretrained model you give,but I only get 0.03962 score in the kaggle.Do you know this is why?I think this is unnormaly!Thank you very much!
Warning is ok. You have zero files in folder. Check if there are some JPGs available.
can you give me some advice to increase the performance?Thank you!
Typical errors:
- Check preprocessing of images
- Check min/max size of images
- Check you read it BGR Also reduce THR for predicted boxes
I propose to check score on validation as well.
@ZFTurbo， I use your code and the model to inference .I think it is not the data probleme ...Thanks！
Typical errors:
- Check preprocessing of images
- Check min/max size of images
- Check you read it BGR Also reduce THR for predicted boxes
I propose to check score on validation as well.
Hi,Excuse me.Thanks for your replay!I use the resnet101 inference model you give,but when I use the inference model to predict the test dataset 2019,the result is not good.Btw,should I use the training model or the inference model to test the 2019 test dataset?And,if I use the 'avg_result' not the 'high_level_relult',I will get a normaly result!So,should I choose any model and result?Thank you!Thank you very much!
As I can see from LB you already solved the problem?
I am also see you in the LB,you use the 2019 train data to retrain the model?Or,you are only use the trained to test the 2019 test dataset?Now,I am not sure the problem is where.But I am not use the file 'create_higher_level_predictions_from_level_1_predictions_csv.py' to expand the result.And I only use the origin result I get,the score is normaly!So I am not sure the problem is where?Btw,Thank you for your project,it helped me a lot!Thanks!
You can download it from here: https://storage.googleapis.com/openimages/v5/class-descriptions-boxable.csv
Hey, @ZFTurbo I'm getting a KeyError when I use this file. Is there any other version I can use?
