![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F3867671%2Fb39bbe97304086f9e034460df1799b66%2F.JPG?generation=1594131634255541&alt=media\")
Hello everyone,\nI submitted my notebook several times and got a score. Today i added some code to my notebook to make my prediction more accurate and then i comitted the notebook. In the output i get a submission file as required (submission.csv):\n
But when i submit the CSV file, i get this:\n![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F3867671%2F652937b383ec989ac0853a72517eb5db%2F2.JPG?generation=1594133057320019&alt=media\")
Any idea why it happens? thanks
", "messages": [ { "id": "918857", "postDate": "07/07/2020 14:44:37", "content": "Hello everyone,\nI submitted my notebook several times and got a score. Today i added some code to my notebook to make my prediction more accurate and then i comitted the notebook. In the output i get a submission file as required (submission.csv):\n
But when i submit the CSV file, i get this:\n
Any idea why it happens? thanks
", "rawMarkdown": "Hello everyone,\nI submitted my notebook several times and got a score. Today i added some code to my notebook to make my prediction more accurate and then i comitted the notebook. In the output i get a submission file as required (submission.csv):\n\n\nBut when i submit the CSV file, i get this:\n\n\n\nAny idea why it happens? thanks", "votes": null }, { "id": "918866", "postDate": "07/07/2020 14:53:20", "content": "https://www.kaggle.com/c/global-wheat-detection/discussion/149871\nhttps://www.kaggle.com/c/global-wheat-detection/discussion/149809\nhttps://www.kaggle.com/c/global-wheat-detection/discussion/149847
", "rawMarkdown": "https://www.kaggle.com/c/global-wheat-detection/discussion/149871\nhttps://www.kaggle.com/c/global-wheat-detection/discussion/149809\nhttps://www.kaggle.com/c/global-wheat-detection/discussion/149847", "votes": null }, { "id": "918877", "postDate": "07/07/2020 14:57:21", "content": "It didn't show what you got. May be your internet got wrong at that time. Not sure about this, but hope can be useful.
", "rawMarkdown": "It didn't show what you got. May be your internet got wrong at that time. Not sure about this, but hope can be useful.", "votes": null }, { "id": "919106", "postDate": "07/07/2020 17:57:41", "content": "i checked them all and my code knows how to deal with images without predictions and the image doesn't has to be 1024x1024
\n\nthe weird about it is that i only changed the prediction function (i added TTA) and since then it happens
", "rawMarkdown": "i checked them all and my code knows how to deal with images without predictions and the image doesn't has to be 1024x1024\n\nthe weird about it is that i only changed the prediction function (i added TTA) and since then it happens", "votes": null }, { "id": "919116", "postDate": "07/07/2020 18:03:32", "content": "Your submission must be crashing at some point while performing inference. Perhaps your TTA assumes that the model generates some box predictions, which may not always be true
", "rawMarkdown": "Your submission must be crashing at some point while performing inference. Perhaps your TTA assumes that the model generates some box predictions, which may not always be true", "votes": null }, { "id": "919495", "postDate": "07/07/2020 23:10:11", "content": "Im using this github to do TTA: https://github.com/user673/TTA-object-detection and it doesn't seem like something is wrong there. Any way i made a function that makes predictions and print the image + the Bboxes according to those predictions and i wrote another function that writes the submit file. When i use the predict_TTA function in both functions (print and submit), the submit fails but when i use it only in the print function and in the submit function i use predict without TTA, it doesn't fails so i don't think it crashes in the predict_TTA part.
", "rawMarkdown": "Im using this github to do TTA: https://github.com/user673/TTA-object-detection and it doesn't seem like something is wrong there. Any way i made a function that makes predictions and print the image + the Bboxes according to those predictions and i wrote another function that writes the submit file. When i use the predict_TTA function in both functions (print and submit), the submit fails but when i use it only in the print function and in the submit function i use predict without TTA, it doesn't fails so i don't think it crashes in the predict_TTA part.", "votes": null }, { "id": "921141", "postDate": "07/09/2020 05:16:13", "content": "Hi @galelias you might forget to add empty string while model is detecting nothing in image.Hope this helps.
", "rawMarkdown": "Hi @galelias you might forget to add empty string while model is detecting nothing in image.Hope this helps.", "votes": null }, { "id": "922907", "postDate": "07/10/2020 12:08:43", "content": "I am having the same issue: used to get score in my notebooks, but today I added new stuff and I get \"Submission CSV Not Found\". However, I have not modified the part of the code that generates the submission CSV. I assume it is crashing in some part of the new code, but nothing seems suspiciousto me
\n\nDid you find where your notebook crashes?
", "rawMarkdown": "I am having the same issue: used to get score in my notebooks, but today I added new stuff and I get \"Submission CSV Not Found\". However, I have not modified the part of the code that generates the submission CSV. I assume it is crashing in some part of the new code, but nothing seems suspiciousto me\n\nDid you find where your notebook crashes?", "votes": null }, { "id": "923194", "postDate": "07/10/2020 15:44:45", "content": "Yes, i figured out that when there's only 1 prediction, \"scores\" and \"labels\" returned as scalar instead of 1d array and bbox returned as 1d array instead of 2d so i used np.atleast_1d and np.atleast_2d to handle this. Moreover, you should check that your model can handle images which are not 1024x1024 and that it stores also images without bboxes as empty string.
", "rawMarkdown": "Yes, i figured out that when there's only 1 prediction, \"scores\" and \"labels\" returned as scalar instead of 1d array and bbox returned as 1d array instead of 2d so i used np.atleast_1d and np.atleast_2d to handle this. Moreover, you should check that your model can handle images which are not 1024x1024 and that it stores also images without bboxes as empty string.", "votes": null }, { "id": "924567", "postDate": "07/11/2020 14:03:35", "content": "I thought I was th empty string problem but it looked ok in my script. Will try the 1 prediction case, never thought about it... thanks!
", "rawMarkdown": "I thought I was th empty string problem but it looked ok in my script. Will try the 1 prediction case, never thought about it... thanks!", "votes": null } ], "comments": [ { "id": 918866, "author_name": "tiandaye", "author_url": "", "post_date": "07/07/2020 14:53:20", "content": "https://www.kaggle.com/c/global-wheat-detection/discussion/149871\nhttps://www.kaggle.com/c/global-wheat-detection/discussion/149809\nhttps://www.kaggle.com/c/global-wheat-detection/discussion/149847
", "votes": null, "replies": [ { "id": 919106, "author_name": "galelias", "author_url": "", "post_date": "07/07/2020 17:57:41", "content": "i checked them all and my code knows how to deal with images without predictions and the image doesn't has to be 1024x1024
\n\nthe weird about it is that i only changed the prediction function (i added TTA) and since then it happens
", "votes": null, "replies": [] }, { "id": 919116, "author_name": "matthewmasters", "author_url": "", "post_date": "07/07/2020 18:03:32", "content": "Your submission must be crashing at some point while performing inference. Perhaps your TTA assumes that the model generates some box predictions, which may not always be true
", "votes": null, "replies": [] } ] }, { "id": 918877, "author_name": "", "author_url": "", "post_date": "07/07/2020 14:57:21", "content": "It didn't show what you got. May be your internet got wrong at that time. Not sure about this, but hope can be useful.
", "votes": null, "replies": [] }, { "id": 919495, "author_name": "galelias", "author_url": "", "post_date": "07/07/2020 23:10:11", "content": "Im using this github to do TTA: https://github.com/user673/TTA-object-detection and it doesn't seem like something is wrong there. Any way i made a function that makes predictions and print the image + the Bboxes according to those predictions and i wrote another function that writes the submit file. When i use the predict_TTA function in both functions (print and submit), the submit fails but when i use it only in the print function and in the submit function i use predict without TTA, it doesn't fails so i don't think it crashes in the predict_TTA part.
", "votes": null, "replies": [] }, { "id": 921141, "author_name": "ravi02516", "author_url": "", "post_date": "07/09/2020 05:16:13", "content": "Hi @galelias you might forget to add empty string while model is detecting nothing in image.Hope this helps.
", "votes": null, "replies": [] }, { "id": 922907, "author_name": "benayas", "author_url": "", "post_date": "07/10/2020 12:08:43", "content": "I am having the same issue: used to get score in my notebooks, but today I added new stuff and I get \"Submission CSV Not Found\". However, I have not modified the part of the code that generates the submission CSV. I assume it is crashing in some part of the new code, but nothing seems suspiciousto me
\n\nDid you find where your notebook crashes?
", "votes": null, "replies": [ { "id": 923194, "author_name": "galelias", "author_url": "", "post_date": "07/10/2020 15:44:45", "content": "Yes, i figured out that when there's only 1 prediction, \"scores\" and \"labels\" returned as scalar instead of 1d array and bbox returned as 1d array instead of 2d so i used np.atleast_1d and np.atleast_2d to handle this. Moreover, you should check that your model can handle images which are not 1024x1024 and that it stores also images without bboxes as empty string.
", "votes": null, "replies": [] }, { "id": 924567, "author_name": "benayas", "author_url": "", "post_date": "07/11/2020 14:03:35", "content": "I thought I was th empty string problem but it looked ok in my script. Will try the 1 prediction case, never thought about it... thanks!
", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "918857": "Hello everyone,\nI submitted my notebook several times and got a score. Today i added some code to my notebook to make my prediction more accurate and then i comitted the notebook. In the output i get a submission file as required (submission.csv):\n\n\nBut when i submit the CSV file, i get this:\n\n\n\nAny idea why it happens? thanks", "918866": "https://www.kaggle.com/c/global-wheat-detection/discussion/149871\nhttps://www.kaggle.com/c/global-wheat-detection/discussion/149809\nhttps://www.kaggle.com/c/global-wheat-detection/discussion/149847", "918877": "It didn't show what you got. May be your internet got wrong at that time. Not sure about this, but hope can be useful.", "919106": "i checked them all and my code knows how to deal with images without predictions and the image doesn't has to be 1024x1024\n\nthe weird about it is that i only changed the prediction function (i added TTA) and since then it happens", "919116": "Your submission must be crashing at some point while performing inference. Perhaps your TTA assumes that the model generates some box predictions, which may not always be true", "919495": "Im using this github to do TTA: https://github.com/user673/TTA-object-detection and it doesn't seem like something is wrong there. Any way i made a function that makes predictions and print the image + the Bboxes according to those predictions and i wrote another function that writes the submit file. When i use the predict_TTA function in both functions (print and submit), the submit fails but when i use it only in the print function and in the submit function i use predict without TTA, it doesn't fails so i don't think it crashes in the predict_TTA part.", "921141": "Hi @galelias you might forget to add empty string while model is detecting nothing in image.Hope this helps.", "922907": "I am having the same issue: used to get score in my notebooks, but today I added new stuff and I get \"Submission CSV Not Found\". However, I have not modified the part of the code that generates the submission CSV. I assume it is crashing in some part of the new code, but nothing seems suspiciousto me\n\nDid you find where your notebook crashes?", "923194": "Yes, i figured out that when there's only 1 prediction, \"scores\" and \"labels\" returned as scalar instead of 1d array and bbox returned as 1d array instead of 2d so i used np.atleast_1d and np.atleast_2d to handle this. Moreover, you should check that your model can handle images which are not 1024x1024 and that it stores also images without bboxes as empty string.", "924567": "I thought I was th empty string problem but it looked ok in my script. Will try the 1 prediction case, never thought about it... thanks!" }, "source": "meta" }