![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F4910466%2F975f817bb51199a44e5bfd530847424a%2FScreen%20Shot%202023-02-04%20at%2017.00.36.png?generation=1675497684223997&alt=media\")
Welcome to the NFL Health and Safety - Player Contact Detection competition!
\nYet again, we are looking forward to seeing how kagglers attack this problem!
\nIf you haven't already, I encourage you to to read the evaluation details, data description and our starter guide notebook.
\nFeel free to post any questions you might have. @samhuddleston, @brittevans and I will try our best to respond promptly.
\nThanks for joining and good luck!
", "messages": [ { "id": 2056202, "postDate": "2022-12-05T21:00:54.257Z", "content": "Welcome to the NFL Health and Safety - Player Contact Detection competition!
\nYet again, we are looking forward to seeing how kagglers attack this problem!
\nIf you haven't already, I encourage you to to read the evaluation details, data description and our starter guide notebook.
\nFeel free to post any questions you might have. @samhuddleston, @brittevans and I will try our best to respond promptly.
\nThanks for joining and good luck!
", "rawMarkdown": "Welcome to the NFL Health and Safety - Player Contact Detection competition!\n\nYet again, we are looking forward to seeing how kagglers attack this problem!\n\nIf you haven't already, I encourage you to to read the [evaluation details](https://www.kaggle.com/competitions/nfl-player-contact-detection/overview/evaluation), [data description](https://www.kaggle.com/competitions/nfl-player-contact-detection/data) and our [starter guide notebook](https://www.kaggle.com/code/robikscube/nfl-player-contact-detection-getting-started).\n\nFeel free to post any questions you might have. @samhuddleston, @brittevans and I will try our best to respond promptly.\n\nThanks for joining and good luck!", "votes": 75 }, { "id": 2059228, "postDate": "2022-12-08T15:46:12.560Z", "content": "@robikscube
\nIn train_baseline_helmets.csv, I find video = 58226_003009_Endzone.mp4 and 58226_003009_Endzone2.mp4.
\nBut in train folder I can only find 58226_003009_Endzone.mp4. What's their difference?
Nice catch @zzy990106 ! This was a mistake on my part. Please ignore any rows in the train_baseline_helmets.csv where the view column value is \"Endzone2\"
Thank you for the clarification.
", "rawMarkdown": "Thank you for the clarification.", "votes": 1 } ] }, { "id": 2056271, "postDate": "2022-12-06T00:06:12.667Z", "content": "@robikscube Thanks for hosting this competition! Hope to see your baseline solution at the twitch live stream
", "rawMarkdown": "@robikscube Thanks for hosting this competition! Hope to see your baseline solution at the twitch live stream ", "votes": 4, "replies": [ { "id": 2057102, "postDate": "2022-12-06T18:40:03.547Z", "content": "I 2nd that ✌️
", "rawMarkdown": "I 2nd that ✌️", "votes": 1 }, { "id": 2057283, "postDate": "2022-12-06T23:52:22.380Z", "content": "Thanks Gabriel! Since I'm helping host this competition I won't be discussing too many details outside the forums. I'm really excited to see what people come up with, and I think this is a super interesting challenge. Feel free to ask any questions here and we will do our best to help.
", "rawMarkdown": "Thanks Gabriel! Since I'm helping host this competition I won't be discussing too many details outside the forums. I'm really excited to see what people come up with, and I think this is a super interesting challenge. Feel free to ask any questions here and we will do our best to help.", "votes": 2 } ] }, { "id": 2154353, "postDate": "2023-02-22T00:47:29.317Z", "content": "May I also ask, are the cameras' positions stationary and only pan/zoom during the game or are cameras moving as well?
", "rawMarkdown": "May I also ask, are the cameras' positions stationary and only pan/zoom during the game or are cameras moving as well?", "votes": 1, "replies": [ { "id": 2156902, "postDate": "2023-02-23T15:52:13.620Z", "content": "@dmytropoplavskiy - The camera position will always be stationary, although the location of the camera relative to the field may differ between stadiums.
", "rawMarkdown": "@dmytropoplavskiy - The camera position will always be stationary, although the location of the camera relative to the field may differ between stadiums.", "votes": 1 } ] }, { "id": 2128963, "postDate": "2023-02-04T08:04:26.877Z", "content": "@robikscube
\nI'm not sure how timestamps are syncted between multiple data.
\nIs this correct picture of timestamps?
start_time (frame=0)snap_time (frame=300)
Another question:
\nIf you are using a 3D or 2.5D CNN, you would need prior image to infer the contact of frame=300.
\nTherefore, I would expect the train_baseline_helmet data to exist from frame 290.
@shinmurashinmura Thanks. I came to understand the merit of frames before snap_time. It also useful to accurately generate sequential features (e.g. diff/shift etc.).
Why do you think snap_time starts at 300 frames? Only the time data is provided and I think the 10 frame difference is due to the way it was calculated during the conversion. Please correct me if you think I am wrong.
", "rawMarkdown": "Why do you think snap_time starts at 300 frames? Only the time data is provided and I think the 10 frame difference is due to the way it was calculated during the conversion. Please correct me if you think I am wrong.", "replies": [ { "id": 2131860, "postDate": "2023-02-06T12:53:32.960Z", "content": "@chenlin1999 Because it is written in the data section.
\n\n", "rawMarkdown": "@chenlin1999 Because it is written in the data section.\n\n> snap_time: The timestamp when the play starts within the video. This is 5 seconds (300 frames) into the video.", "votes": 1 }, { "id": 2132487, "postDate": "2023-02-06T20:51:49.943Z", "content": "snap_time: The timestamp when the play starts within the video. This is 5 seconds (300 frames) into the video.
\n
Yes, you are right here, snap_time is always 5s after the start of the video. This means that they always intercept the video 5s more before the start of the game. but your graph above I think is incorrect because tracker_data doesn't start at snap_time, it starts much earlier. As in gameplay '58580_001136', tracker_data starts 10.8s earlier, there is not even video at this point, you can look at the step column, it starts at -108. step=0 is exactly snap_time. so I don't think it is 290-299 frame has no data, it's just that helmetlabel is also 10 frames ahead.
\nPlease correct me if you think I am wrong.
", "rawMarkdown": "Yes, you are right here, snap_time is always 5s after the start of the video. This means that they always intercept the video 5s more before the start of the game. but your graph above I think is incorrect because tracker_data doesn't start at snap_time, it starts much earlier. As in gameplay '58580_001136', tracker_data starts 10.8s earlier, there is not even video at this point, you can look at the step column, it starts at -108. step=0 is exactly snap_time. so I don't think it is 290-299 frame has no data, it's just that helmetlabel is also 10 frames ahead.\n\nPlease correct me if you think I am wrong.", "votes": 1 }, { "id": 2132593, "postDate": "2023-02-06T22:33:20.137Z", "content": "@chenlin1999 Oh, I understand your point. Yes, since tracker data is started before snap_time, the arrow of tracker data should be extended before snap_time. It's just a mistake. Thank you for pointing out.
Sorry for the delayed response @tatamikenn - Your drawing looks correct to me. The only thing I'd note is that the snap_frame is approximate and could potentially be off by a few frames. The tracking data is also sampled at a different rate (10Hz) so you may need to experiment with different approaches if joining the two.
We've provided helmets predictions for 10 frames prior to the snap as a buffer.
\nHope that helps!
", "rawMarkdown": "Sorry for the delayed response @tatamikenn - Your drawing looks correct to me. The only thing I'd note is that the `snap_frame` is approximate and could potentially be off by a few frames. The tracking data is also sampled at a different rate (10Hz) so you may need to experiment with different approaches if joining the two.\n\nWe've provided helmets predictions for 10 frames prior to the snap as a buffer.\n\nHope that helps!", "votes": 1 } ] }, { "id": 2124315, "postDate": "2023-01-31T23:36:11.477Z", "content": "where is the test.csv? I mean I build the model,but there is no test.csv.Is the testset data need to split from train dataset?
", "rawMarkdown": "where is the test.csv? I mean I build the model,but there is no test.csv.Is the testset data need to split from train dataset?", "votes": 1, "replies": [ { "id": 2133598, "postDate": "2023-02-07T14:08:19.173Z", "content": "I'd suggest reading through the competition data description here: https://www.kaggle.com/competitions/nfl-player-contact-detection/data
\nThere are three test related csv files:
\nThis is a code competition. When you submit, your model will be rerun on a set of 61 unseen plays located in a holdout test set. The publicly provided test videos are simply a set of mock plays (copied from the training set) which are not used in scoring.\n\nThe associated test_baseline_helmets.csv, test_player_tracking.csv, and test_video_metadata.csv are available to your model when submitting.\n@robikscube
\nI have concerns about the player's tracking dataset. The step feature has a max value of 692 which means those videos should have 69.2 seconds of duration! Of the videos, we have the longest one has 22.15 seconds duration. Why is this happening? Also, I think I understand this, but there are negative steps. I reason that they are before the game starts and from 0 starts the game. I would appreciate some explanation!
Thanks for the question @dyordanov - Good find in noticing the large step values. Keep in mind that the tracking data may include time before and after the duration of the play and outside the time of the video. You are only required to predict contact that occurs for the steps in the sample_submission.csv file, which are timesteps where there is both sideline and endzone video. The same is true for the provided labels in the training set.
As you mentioned some plays have large step values in the tracking data:
>> df = pd.read_csv('../input/nfl-player-contact-detection/train_player_tracking.csv')\n>> df.groupby('game_play')['step'].max().sort_values(ascending=False)\n\ngame_play\n58225_003821 692\n58491_000084 593\n58348_003069 527\n58509_003605 348\n58520_002380 311\n...\nHowever the max step in the labels are much smaller:
>> labels = pd.read_csv('../input/nfl-player-contact-detection/train_labels.csv')\n>> labels.groupby('game_play')['step'].max().sort_values(ascending=False)\n\ngame_play\n58537_000757 172\n58524_002909 163\n58224_003139 134\n58173_003606 129\n58172_003247 125\n...\nIn short, there is a lot of extra tracking data after the action of the plays that you can choose to ignore.
\nHope that helps.
", "rawMarkdown": "Thanks for the question @dyordanov - Good find in noticing the large step values. Keep in mind that the tracking data may include time before and after the duration of the play and outside the time of the video. You are only required to predict contact that occurs for the steps in the `sample_submission.csv` file, which are timesteps where there is both sideline and endzone video. The same is true for the provided labels in the training set.\n\nAs you mentioned some plays have large `step` values in the tracking data:\n```\n>> df = pd.read_csv('../input/nfl-player-contact-detection/train_player_tracking.csv')\n>> df.groupby('game_play')['step'].max().sort_values(ascending=False)\n\ngame_play\n58225_003821 692\n58491_000084 593\n58348_003069 527\n58509_003605 348\n58520_002380 311\n...\n```\n\nHowever the max `step` in the labels are much smaller:\n\n```\n>> labels = pd.read_csv('../input/nfl-player-contact-detection/train_labels.csv')\n>> labels.groupby('game_play')['step'].max().sort_values(ascending=False)\n\ngame_play\n58537_000757 172\n58524_002909 163\n58224_003139 134\n58173_003606 129\n58172_003247 125\n...\n```\n\nIn short, there is a lot of extra tracking data after the action of the plays that you can choose to ignore.\n\nHope that helps.", "votes": 2, "replies": [ { "id": 2102194, "postDate": "2023-01-16T13:01:16.370Z", "content": "Yes, this explains things! Thank you!
", "rawMarkdown": "Yes, this explains things! Thank you!", "votes": 1 } ] } ] }, { "id": 2100361, "postDate": "2023-01-15T05:16:50.003Z", "content": "Hey Rob, can we get the baseline helmets for All29?
", "rawMarkdown": "Hey Rob, can we get the baseline helmets for All29?", "votes": 1, "replies": [ { "id": 2101347, "postDate": "2023-01-15T21:11:27.707Z", "content": "Unfortunately we do not have the baseline helmets for the All29 views.
", "rawMarkdown": "Unfortunately we do not have the baseline helmets for the All29 views.", "votes": 1 } ] }, { "id": 2067191, "postDate": "2022-12-16T13:43:29.547Z", "content": "@robikscube
\nThe dataset description page states that there is a column called \"event\" in [train/test]_player_tracking.csv, but the actual data does not contain it. Is this a typographical error in the description?
Yes, this is a typo. Events are not included in the tracking data for this competition. Thanks for mentioning it. I will remove it from the data description.
", "rawMarkdown": "Yes, this is a typo. Events are not included in the tracking data for this competition. Thanks for mentioning it. I will remove it from the data description.", "votes": 1, "replies": [ { "id": 2067201, "postDate": "2022-12-16T13:52:42.430Z", "content": "OK. Thanks for the quick fix!
", "rawMarkdown": "OK. Thanks for the quick fix!", "votes": 1 } ] } ] }, { "id": 2065288, "postDate": "2022-12-14T13:50:36.937Z", "content": "@robikscube
\nRegarding \"Submission Limits\" in Rules section, is the current description correct?
\n", "rawMarkdown": "@robikscube \nRegarding \"Submission Limits\" in Rules section, is the current description correct? \n> There is no daily submission limit. Only the most recent submission by a team will be evaluated after the challenge deadline.", "votes": 1, "replies": [ { "id": 2065347, "postDate": "2022-12-14T15:03:44.033Z", "content": "There is no daily submission limit. Only the most recent submission by a team will be evaluated after the challenge deadline.
\n
Fixed that to 5! Sorry, that was old copy from an analytics competition.
", "rawMarkdown": "Fixed that to 5! Sorry, that was old copy from an analytics competition.", "votes": 1 }, { "id": 2065350, "postDate": "2022-12-14T15:07:51.010Z", "content": "Thanks for the quick fix!
", "rawMarkdown": "Thanks for the quick fix!" } ] }, { "id": 2058271, "postDate": "2022-12-07T18:52:20.433Z", "content": "How Exciting!
", "rawMarkdown": "How Exciting!", "votes": 1 }, { "id": 2056749, "postDate": "2022-12-06T12:08:36.327Z", "content": "It is great opportunity to update knowledge. Let's move.
", "rawMarkdown": "It is great opportunity to update knowledge. Let's move.", "votes": 1 }, { "id": 2132995, "postDate": "2023-02-07T07:22:13.927Z", "content": "I'm new on kaggle i don't know how to do ,here in data many csv files which one we have to use and can i start on google colab ?
", "rawMarkdown": "I'm new on kaggle i don't know how to do ,here in data many csv files which one we have to use and can i start on google colab ?", "votes": 2, "replies": [ { "id": 2133592, "postDate": "2023-02-07T14:05:00.347Z", "content": "Welcome! I'd suggest checking out some of the existing starter notebooks here on kaggle and working in a kaggle notebook. That will ensure you have access to the competition data.
", "rawMarkdown": "Welcome! I'd suggest checking out some of the existing starter notebooks here on kaggle and working in a kaggle notebook. That will ensure you have access to the competition data.", "votes": 1, "replies": [ { "id": 2134660, "postDate": "2023-02-08T07:03:01.247Z", "content": "Thank you for valuable response.
", "rawMarkdown": "Thank you for valuable response.", "votes": 1 }, { "id": 2135276, "postDate": "2023-02-08T14:37:15.910Z", "content": "\n\nWelcome! I'd suggest checking out some of the existing starter notebooks here on kaggle and working in a kaggle notebook. That will ensure you have access to the competition data.
\n
cool
", "rawMarkdown": "> Welcome! I'd suggest checking out some of the existing starter notebooks here on kaggle and working in a kaggle notebook. That will ensure you have access to the competition data.\n\ncool\n" }, { "id": 2138128, "postDate": "2023-02-10T15:26:06.583Z", "content": "yes,thank you
", "rawMarkdown": "yes,thank you\n" } ] } ] }, { "id": 2115065, "postDate": "2023-01-25T13:02:02.237Z", "content": "@robikscube quick questions - not sure if it was already covered somewhere:
\nGreat questions.
\nHope that answers your questions!
", "rawMarkdown": "Great questions.\n- Yes, using publicly available data is permitted. However, hand labeling or attempting to identify the **test set** plays is not permitted.\n- Yolo models are approved. We are working on getting the wording modified in the official rules to state \"Entries leveraging tools subject to the MIT, GNU GPL 3.0, and Apache 2.0 open source licenses are approved for use in the competition\", similar to what it has stated in prior years.\n\nHope that answers your questions!", "votes": 7, "replies": [ { "id": 2116580, "postDate": "2023-01-26T15:21:20.620Z", "content": "\n\nYes, using publicly available data is permitted. However, hand labeling or attempting to identify the test set plays is not permitted.
\n
Do you define old Kaggle competition data as publicly available?
\nAny hand labeling of training data or publicly available data is not allowed?
", "rawMarkdown": "> Yes, using publicly available data is permitted. However, hand labeling or attempting to identify the test set plays is not permitted.\n\nDo you define old Kaggle competition data as publicly available?\n\nAny hand labeling of training data or publicly available data is not allowed?", "votes": 1 }, { "id": 2116848, "postDate": "2023-01-26T19:19:16.240Z", "content": "So long as it meets the below criteria, external data may be used. Last year's competition data is still publically available, so we would consider it covered as permitted for use.
\nC. External Data. You may use data other than the Competition Data (“External Data”) to develop and test your Submissions. However, you will ensure the External Data is publicly available to use by all participants of the Competition for purposes of the competition at no cost to the other participants. The ability to use External Data under this Section 7.C (External Data) does not limit you other obligations under these Competition Rules, including but not limited to Section 11 (Winners Obligations).\nLabeling or pseudo-labeling of training or external data is fine, the distinction being that it is against the rules to attempt to hand label the test data. Per the rules:
\nSubmissions may not use or incorporate information from hand labeling or human prediction of the validation dataset or test data records.\nBecause this is a code competition, the public and private test set is unknown. Any attempt to identify and/or label the test plays would be considered against the spirit of the competition.
\nHope that helps. If you care to elaborate more on how you intend to use external data, I could clarify more specifically.
", "rawMarkdown": "So long as it meets the below criteria, external data may be used. Last year's competition data is still publically available, so we would consider it covered as permitted for use.\n\n```\nC. External Data. You may use data other than the Competition Data (“External Data”) to develop and test your Submissions. However, you will ensure the External Data is publicly available to use by all participants of the Competition for purposes of the competition at no cost to the other participants. The ability to use External Data under this Section 7.C (External Data) does not limit you other obligations under these Competition Rules, including but not limited to Section 11 (Winners Obligations).\n```\n\nLabeling or pseudo-labeling of training or external data is fine, the distinction being that it is against the rules to attempt to hand label the **test** data. Per the rules:\n\n```\nSubmissions may not use or incorporate information from hand labeling or human prediction of the validation dataset or test data records.\n```\n\nBecause this is a code competition, the public and private test set is unknown. Any attempt to identify and/or label the test plays would be considered against the spirit of the competition.\n\nHope that helps. If you care to elaborate more on how you intend to use external data, I could clarify more specifically.\n", "votes": 3 }, { "id": 2120256, "postDate": "2023-01-29T13:11:36.913Z", "content": "Hi @robikscube,
\nHow about other Kaggle competitons' dataset, like DFL? It is allowed to use?
\nhttps://www.kaggle.com/competitions/dfl-bundesliga-data-shootout
@bamps53 - The DFL competition rules state that the dataset provided should only be used for the specific competition and should not be used for any other purpose. Because of those restrictions I would suggest you not use it in this competition.
\nhttps://www.kaggle.com/competitions/dfl-bundesliga-data-shootout/rules
\nSection 2.7.A: \nA. Data Access and Use. You may access and use the Competition Data only for participating in the Competition and on Kaggle.com forums...\n@robikscube The rules for older NFL competitions also have a similar statement to the DFL competition
\nhttps://www.kaggle.com/competitions/nfl-impact-detection/rules
\n\n\nParticipants in the Data Science Theme may access and use the Competition Data only for participating in the Data Science Theme and on Kaggle.com forums. The Sponsor and Named Partners reserve the right to disqualify any such Participant who uses the Competition Data other than as permitted by the Data Science Theme and these Official Rules. For the avoidance of doubt, Competition Data may include video footage from certain NFL games, and all such footage may be accessed and used only for participating in the Data Science Theme.
\n
Based on my understanding of the rules and your comment above, data from older NFL competitions cannot be used for training our models. Could you please confirm this?
", "rawMarkdown": "@robikscube The rules for older NFL competitions also have a similar statement to the DFL competition\n\nhttps://www.kaggle.com/competitions/nfl-impact-detection/rules\n\n> Participants in the Data Science Theme may access and use the Competition Data only for participating in the Data Science Theme and on Kaggle.com forums. The Sponsor and Named Partners reserve the right to disqualify any such Participant who uses the Competition Data other than as permitted by the Data Science Theme and these Official Rules. For the avoidance of doubt, Competition Data may include video footage from certain NFL games, and all such footage may be accessed and used only for participating in the Data Science Theme.\n\nBased on my understanding of the rules and your comment above, data from older NFL competitions cannot be used for training our models. Could you please confirm this?", "votes": 1 }, { "id": 2133582, "postDate": "2023-02-07T13:49:24.600Z", "content": "Good catch. Considering this competition is a continuation of the NFL's \"First and Future\" efforts, the use of the prior \"First and Future\" competition data for this competition is ok.
", "rawMarkdown": "Good catch. Considering this competition is a continuation of the NFL's \"First and Future\" efforts, the use of the prior \"First and Future\" competition data for this competition is ok.", "votes": 1 } ] } ] }, { "id": 2127371, "postDate": "2023-02-02T20:53:35.617Z", "content": "How Exciting!
", "rawMarkdown": "How Exciting!\n\n\n" }, { "id": 2059094, "postDate": "2022-12-08T13:30:55.607Z", "content": "so exciting go eaglesss
", "rawMarkdown": "so exciting go eaglesss", "votes": 2 }, { "id": 2058629, "postDate": "2022-12-08T04:59:04.363Z", "content": "@robikscube
\nI've a question about the price money for 1st place to 5th place. I think it is unfair to offer the 1st place so much (50%) credit.
1st Place - $50,000\n2nd Place - $25,000\n3rd Place - $13,000\n4th Place - $7,000\n5th Place - $5,000\nAs we see that, normally, the scores between top places don't vary much, so there is no such that level of nobility for particular top solutions. I am cordially suggesting to distribute the price money among top 10 for this high budget competition.
", "rawMarkdown": "@robikscube \nI've a question about the price money for 1st place to 5th place. I think it is unfair to offer the 1st place so much (50%) credit. \n\n```\n1st Place - $50,000\n2nd Place - $25,000\n3rd Place - $13,000\n4th Place - $7,000\n5th Place - $5,000\n```\n\nAs we see that, normally, the scores between top places don't vary much, so there is no such **that level of nobility** for particular top solutions. I am cordially suggesting to distribute the price money among **top 10** for this high budget competition. ", "replies": [ { "id": 2059178, "postDate": "2022-12-08T14:34:52.310Z", "content": "Hey @simonalerdic - I apprecaite your feedback! You make an interesting point. We may consider it for future competitions. The prize structure for this competition is set and we will not be changing it. Good luck!
", "rawMarkdown": "Hey @simonalerdic - I apprecaite your feedback! You make an interesting point. We may consider it for future competitions. The prize structure for this competition is set and we will not be changing it. Good luck!", "votes": 1 }, { "id": 2091253, "postDate": "2023-01-08T07:12:30.370Z", "content": "If you checked last year's NFL competition: https://www.kaggle.com/competitions/nfl-health-and-safety-helmet-assignment/leaderboard
\nand last last year's NFL competition: https://www.kaggle.com/competitions/nfl-impact-detection/leaderboard
\nyou wouldn't post this…
let us do the best we can in this competition as the intention has always been to test our skills and continuous improvement!
", "rawMarkdown": "let us do the best we can in this competition as the intention has always been to test our skills and continuous improvement!", "isDeleted": true } ] }, { "id": 2154460, "postDate": "2023-02-22T03:04:53.207Z", "content": "How Exciting!
", "rawMarkdown": "How Exciting!", "votes": -1 }, { "id": 2718145, "postDate": "2024-03-27T01:57:45.730Z", "content": "Hello, I participated in kaggle's 1st and Future - Player Contact Detection competition and achieved top8%, a bronze medal score of 66/939. Now I need a certificate to prove my score. Can you tell me where I can get it? Have you obtained this certification? Thank you very much.
", "rawMarkdown": "Hello, I participated in kaggle's 1st and Future - Player Contact Detection competition and achieved top8%, a bronze medal score of 66/939. Now I need a certificate to prove my score. Can you tell me where I can get it? Have you obtained this certification? Thank you very much." }, { "id": 2165124, "postDate": "2023-03-02T02:21:22.917Z", "content": "Exciting and interesting!
", "rawMarkdown": "Exciting and interesting!" }, { "id": 2159345, "postDate": "2023-02-25T17:34:41.937Z", "content": "Great competition!
", "rawMarkdown": "Great competition!" }, { "id": 2152476, "postDate": "2023-02-20T20:15:40.363Z", "content": "May I ask where the sensor that records tracking data is located on the player? Is it in the helmet? Or is it on the chest? Or are the different data coming from multiple sensors on multiple parts of the body?
", "rawMarkdown": "May I ask where the sensor that records tracking data is located on the player? Is it in the helmet? Or is it on the chest? Or are the different data coming from multiple sensors on multiple parts of the body?", "replies": [ { "id": 2152519, "postDate": "2023-02-20T20:58:47.543Z", "content": "I believe the sensors are in their shoulder pads, but that may be old information (I see articles on google from 2015-2020 about it, but no newer ones)
", "rawMarkdown": "I believe the sensors are in their shoulder pads, but that may be old information (I see articles on google from 2015-2020 about it, but no newer ones)", "votes": 1, "replies": [ { "id": 2152612, "postDate": "2023-02-20T22:47:20.620Z", "content": "Do you know then approximately what the furthest distance between the two shoulders is if two players are in foot to foot contact? I'm not sure how the height should be converted to the distance units here.
", "rawMarkdown": "Do you know then approximately what the furthest distance between the two shoulders is if two players are in foot to foot contact? I'm not sure how the height should be converted to the distance units here." } ] }, { "id": 2153972, "postDate": "2023-02-21T18:36:51.300Z", "content": "@chenlin1999 - To confirm the answer to your question about sensor locations, you can refer to this: https://operations.nfl.com/gameday/technology/nfl-next-gen-stats/
\n\"2–3 radio-frequency identification (RFID) tags installed into the players’ shoulder pads\"
Thank you!
", "rawMarkdown": "Thank you!" } ] } ] }, { "id": 2151281, "postDate": "2023-02-20T01:44:16.840Z", "content": "I just realized there are some helmets missed for both sideline and endzone views (may because they are too small in the sideline view) that are postive contacts in the label csv. For example, game_play = '58173_003606' and frame = 710-716. Will the test data have the similar cases?
", "rawMarkdown": "I just realized there are some helmets missed for both sideline and endzone views (may because they are too small in the sideline view) that are postive contacts in the label csv. For example, game_play = '58173_003606' and frame = 710-716. Will the test data have the similar cases?", "replies": [ { "id": 2156898, "postDate": "2023-02-23T15:50:18.457Z", "content": "@annealer - yes, you can assume the test data is similar in nature to the training dataset.
", "rawMarkdown": "@annealer - yes, you can assume the test data is similar in nature to the training dataset." } ] }, { "id": 2143939, "postDate": "2023-02-14T16:37:22.127Z", "content": "its exciting.
", "rawMarkdown": "its exciting." }, { "id": 2135270, "postDate": "2023-02-08T14:35:01.870Z", "content": "I'm new to kaggle and this project feels really interesting! How do I get started?？
", "rawMarkdown": "I'm new to kaggle and this project feels really interesting! How do I get started?？" }, { "id": 2092884, "postDate": "2023-01-09T17:12:02.227Z", "content": "Looks interesting to try!
", "rawMarkdown": "Looks interesting to try!" }, { "id": 2090436, "postDate": "2023-01-07T10:34:50.577Z", "content": "Good one and very really interesting
", "rawMarkdown": "Good one and very really interesting " }, { "id": 2150548, "postDate": "2023-02-19T11:08:10.103Z", "rawMarkdown": "", "isDeleted": true }, { "id": 2143835, "postDate": "2023-02-14T14:39:25.507Z", "rawMarkdown": "", "isDeleted": true }, { "id": 2068529, "postDate": "2022-12-18T04:54:12.177Z", "rawMarkdown": "", "isDeleted": true, "replies": [ { "id": 2069367, "postDate": "2022-12-18T21:50:16.470Z", "content": "@palantirha - \"frame\" refers to a still image which makes up the video: https://en.wikipedia.org/wiki/Film_frame
\nHave you read the data description? You might find it helpful. https://www.kaggle.com/competitions/nfl-player-contact-detection/data
", "rawMarkdown": "@palantirha - \"frame\" refers to a still image which makes up the video: https://en.wikipedia.org/wiki/Film_frame\n\nHave you read the data description? You might find it helpful. https://www.kaggle.com/competitions/nfl-player-contact-detection/data", "replies": [ { "id": 2069906, "postDate": "2022-12-19T12:33:42.480Z", "rawMarkdown": "", "votes": 1, "isDeleted": true } ] } ] }, { "id": 2062465, "postDate": "2022-12-12T05:56:35.570Z", "rawMarkdown": "", "votes": 2, "isDeleted": true }, { "id": 2101875, "postDate": "2023-01-16T08:49:24.597Z", "content": "Thank u for hosting
", "rawMarkdown": "Thank u for hosting", "votes": 1 }, { "id": 2061239, "postDate": "2022-12-10T21:27:49.140Z", "content": "**Thanks a lot!! **@robikscube
", "rawMarkdown": "**Thanks a lot!! **@robikscube ", "votes": 1 } ], "comments": [ { "id": 2059228, "author_name": "Leon", "author_url": "", "post_date": "2022-12-08T15:46:12.560000", "content": "@robikscube
\nIn train_baseline_helmets.csv, I find video = 58226_003009_Endzone.mp4 and 58226_003009_Endzone2.mp4.
\nBut in train folder I can only find 58226_003009_Endzone.mp4. What's their difference?
Nice catch @zzy990106 ! This was a mistake on my part. Please ignore any rows in the train_baseline_helmets.csv where the view column value is \"Endzone2\"
Thank you for the clarification.
", "votes": 1, "replies": [] } ] }, { "id": 2056271, "author_name": "Gabriel Lins", "author_url": "", "post_date": "2022-12-06T00:06:12.667000", "content": "@robikscube Thanks for hosting this competition! Hope to see your baseline solution at the twitch live stream
", "votes": 4, "replies": [ { "id": 2057102, "author_name": "Ahmed Eltom", "author_url": "", "post_date": "2022-12-06T18:40:03.547000", "content": "I 2nd that ✌️
", "votes": 1, "replies": [] }, { "id": 2057283, "author_name": "Rob Mulla", "author_url": "", "post_date": "2022-12-06T23:52:22.380000", "content": "Thanks Gabriel! Since I'm helping host this competition I won't be discussing too many details outside the forums. I'm really excited to see what people come up with, and I think this is a super interesting challenge. Feel free to ask any questions here and we will do our best to help.
", "votes": 2, "replies": [] } ] }, { "id": 2154353, "author_name": "Dmytro Poplavskiy", "author_url": "", "post_date": "2023-02-22T00:47:29.317000", "content": "May I also ask, are the cameras' positions stationary and only pan/zoom during the game or are cameras moving as well?
", "votes": 1, "replies": [ { "id": 2156902, "author_name": "Rob Mulla", "author_url": "", "post_date": "2023-02-23T15:52:13.620000", "content": "@dmytropoplavskiy - The camera position will always be stationary, although the location of the camera relative to the field may differ between stadiums.
", "votes": 1, "replies": [] } ] }, { "id": 2128963, "author_name": "Bilzard", "author_url": "", "post_date": "2023-02-04T08:04:26.877000", "content": "@robikscube
\nI'm not sure how timestamps are syncted between multiple data.
\nIs this correct picture of timestamps?
start_time (frame=0)snap_time (frame=300)
Another question:
\nIf you are using a 3D or 2.5D CNN, you would need prior image to infer the contact of frame=300.
\nTherefore, I would expect the train_baseline_helmet data to exist from frame 290.
@shinmurashinmura Thanks. I came to understand the merit of frames before snap_time. It also useful to accurately generate sequential features (e.g. diff/shift etc.).
Why do you think snap_time starts at 300 frames? Only the time data is provided and I think the 10 frame difference is due to the way it was calculated during the conversion. Please correct me if you think I am wrong.
", "votes": 0, "replies": [ { "id": 2131860, "author_name": "Bilzard", "author_url": "", "post_date": "2023-02-06T12:53:32.960000", "content": "@chenlin1999 Because it is written in the data section.
\n\n", "votes": 1, "replies": [] }, { "id": 2132487, "author_name": "Chen Lin", "author_url": "", "post_date": "2023-02-06T20:51:49.943000", "content": "snap_time: The timestamp when the play starts within the video. This is 5 seconds (300 frames) into the video.
\n
Yes, you are right here, snap_time is always 5s after the start of the video. This means that they always intercept the video 5s more before the start of the game. but your graph above I think is incorrect because tracker_data doesn't start at snap_time, it starts much earlier. As in gameplay '58580_001136', tracker_data starts 10.8s earlier, there is not even video at this point, you can look at the step column, it starts at -108. step=0 is exactly snap_time. so I don't think it is 290-299 frame has no data, it's just that helmetlabel is also 10 frames ahead.
\nPlease correct me if you think I am wrong.
", "votes": 1, "replies": [] }, { "id": 2132593, "author_name": "Bilzard", "author_url": "", "post_date": "2023-02-06T22:33:20.137000", "content": "@chenlin1999 Oh, I understand your point. Yes, since tracker data is started before snap_time, the arrow of tracker data should be extended before snap_time. It's just a mistake. Thank you for pointing out.
Sorry for the delayed response @tatamikenn - Your drawing looks correct to me. The only thing I'd note is that the snap_frame is approximate and could potentially be off by a few frames. The tracking data is also sampled at a different rate (10Hz) so you may need to experiment with different approaches if joining the two.
We've provided helmets predictions for 10 frames prior to the snap as a buffer.
\nHope that helps!
", "votes": 1, "replies": [] } ] }, { "id": 2124315, "author_name": "charleschen1", "author_url": "", "post_date": "2023-01-31T23:36:11.477000", "content": "where is the test.csv? I mean I build the model,but there is no test.csv.Is the testset data need to split from train dataset?
", "votes": 1, "replies": [ { "id": 2133598, "author_name": "Rob Mulla", "author_url": "", "post_date": "2023-02-07T14:08:19.173000", "content": "I'd suggest reading through the competition data description here: https://www.kaggle.com/competitions/nfl-player-contact-detection/data
\nThere are three test related csv files:
\nThis is a code competition. When you submit, your model will be rerun on a set of 61 unseen plays located in a holdout test set. The publicly provided test videos are simply a set of mock plays (copied from the training set) which are not used in scoring.\n\nThe associated test_baseline_helmets.csv, test_player_tracking.csv, and test_video_metadata.csv are available to your model when submitting.\n@robikscube
\nI have concerns about the player's tracking dataset. The step feature has a max value of 692 which means those videos should have 69.2 seconds of duration! Of the videos, we have the longest one has 22.15 seconds duration. Why is this happening? Also, I think I understand this, but there are negative steps. I reason that they are before the game starts and from 0 starts the game. I would appreciate some explanation!
Thanks for the question @dyordanov - Good find in noticing the large step values. Keep in mind that the tracking data may include time before and after the duration of the play and outside the time of the video. You are only required to predict contact that occurs for the steps in the sample_submission.csv file, which are timesteps where there is both sideline and endzone video. The same is true for the provided labels in the training set.
As you mentioned some plays have large step values in the tracking data:
>> df = pd.read_csv('../input/nfl-player-contact-detection/train_player_tracking.csv')\n>> df.groupby('game_play')['step'].max().sort_values(ascending=False)\n\ngame_play\n58225_003821 692\n58491_000084 593\n58348_003069 527\n58509_003605 348\n58520_002380 311\n...\nHowever the max step in the labels are much smaller:
>> labels = pd.read_csv('../input/nfl-player-contact-detection/train_labels.csv')\n>> labels.groupby('game_play')['step'].max().sort_values(ascending=False)\n\ngame_play\n58537_000757 172\n58524_002909 163\n58224_003139 134\n58173_003606 129\n58172_003247 125\n...\nIn short, there is a lot of extra tracking data after the action of the plays that you can choose to ignore.
\nHope that helps.
", "votes": 2, "replies": [ { "id": 2102194, "author_name": "Daniel Yordanov", "author_url": "", "post_date": "2023-01-16T13:01:16.370000", "content": "Yes, this explains things! Thank you!
", "votes": 1, "replies": [] } ] } ] }, { "id": 2100361, "author_name": "arutema47", "author_url": "", "post_date": "2023-01-15T05:16:50.003000", "content": "Hey Rob, can we get the baseline helmets for All29?
", "votes": 1, "replies": [ { "id": 2101347, "author_name": "Rob Mulla", "author_url": "", "post_date": "2023-01-15T21:11:27.707000", "content": "Unfortunately we do not have the baseline helmets for the All29 views.
", "votes": 1, "replies": [] } ] }, { "id": 2067191, "author_name": "penguin46", "author_url": "", "post_date": "2022-12-16T13:43:29.547000", "content": "@robikscube
\nThe dataset description page states that there is a column called \"event\" in [train/test]_player_tracking.csv, but the actual data does not contain it. Is this a typographical error in the description?
Yes, this is a typo. Events are not included in the tracking data for this competition. Thanks for mentioning it. I will remove it from the data description.
", "votes": 1, "replies": [ { "id": 2067201, "author_name": "penguin46", "author_url": "", "post_date": "2022-12-16T13:52:42.430000", "content": "OK. Thanks for the quick fix!
", "votes": 1, "replies": [] } ] } ] }, { "id": 2065288, "author_name": "kzykmyzw", "author_url": "", "post_date": "2022-12-14T13:50:36.937000", "content": "@robikscube
\nRegarding \"Submission Limits\" in Rules section, is the current description correct?
\n", "votes": 1, "replies": [ { "id": 2065347, "author_name": "Addison Howard", "author_url": "", "post_date": "2022-12-14T15:03:44.033000", "content": "There is no daily submission limit. Only the most recent submission by a team will be evaluated after the challenge deadline.
\n
Fixed that to 5! Sorry, that was old copy from an analytics competition.
", "votes": 1, "replies": [] }, { "id": 2065350, "author_name": "kzykmyzw", "author_url": "", "post_date": "2022-12-14T15:07:51.010000", "content": "Thanks for the quick fix!
", "votes": 0, "replies": [] } ] }, { "id": 2058271, "author_name": "Emma Xuejun Sui", "author_url": "", "post_date": "2022-12-07T18:52:20.433000", "content": "How Exciting!
", "votes": 1, "replies": [] }, { "id": 2056749, "author_name": "Dr. Suhas Rokde", "author_url": "", "post_date": "2022-12-06T12:08:36.327000", "content": "It is great opportunity to update knowledge. Let's move.
", "votes": 1, "replies": [] }, { "id": 2132995, "author_name": "Bhagyashri Dhavale ", "author_url": "", "post_date": "2023-02-07T07:22:13.927000", "content": "I'm new on kaggle i don't know how to do ,here in data many csv files which one we have to use and can i start on google colab ?
", "votes": 2, "replies": [ { "id": 2133592, "author_name": "Rob Mulla", "author_url": "", "post_date": "2023-02-07T14:05:00.347000", "content": "Welcome! I'd suggest checking out some of the existing starter notebooks here on kaggle and working in a kaggle notebook. That will ensure you have access to the competition data.
", "votes": 1, "replies": [ { "id": 2134660, "author_name": "Bhagyashri Dhavale ", "author_url": "", "post_date": "2023-02-08T07:03:01.247000", "content": "Thank you for valuable response.
", "votes": 1, "replies": [] }, { "id": 2135276, "author_name": "chenkeshuai", "author_url": "", "post_date": "2023-02-08T14:37:15.910000", "content": "\n\nWelcome! I'd suggest checking out some of the existing starter notebooks here on kaggle and working in a kaggle notebook. That will ensure you have access to the competition data.
\n
cool
", "votes": 0, "replies": [] }, { "id": 2138128, "author_name": "Bhagyashri Dhavale ", "author_url": "", "post_date": "2023-02-10T15:26:06.583000", "content": "yes,thank you
", "votes": 0, "replies": [] } ] } ] }, { "id": 2115065, "author_name": "Psi", "author_url": "", "post_date": "2023-01-25T13:02:02.237000", "content": "@robikscube quick questions - not sure if it was already covered somewhere:
\nGreat questions.
\nHope that answers your questions!
", "votes": 7, "replies": [ { "id": 2116580, "author_name": "Psi", "author_url": "", "post_date": "2023-01-26T15:21:20.620000", "content": "\n\nYes, using publicly available data is permitted. However, hand labeling or attempting to identify the test set plays is not permitted.
\n
Do you define old Kaggle competition data as publicly available?
\nAny hand labeling of training data or publicly available data is not allowed?
", "votes": 1, "replies": [] }, { "id": 2116848, "author_name": "Rob Mulla", "author_url": "", "post_date": "2023-01-26T19:19:16.240000", "content": "So long as it meets the below criteria, external data may be used. Last year's competition data is still publically available, so we would consider it covered as permitted for use.
\nC. External Data. You may use data other than the Competition Data (“External Data”) to develop and test your Submissions. However, you will ensure the External Data is publicly available to use by all participants of the Competition for purposes of the competition at no cost to the other participants. The ability to use External Data under this Section 7.C (External Data) does not limit you other obligations under these Competition Rules, including but not limited to Section 11 (Winners Obligations).\nLabeling or pseudo-labeling of training or external data is fine, the distinction being that it is against the rules to attempt to hand label the test data. Per the rules:
\nSubmissions may not use or incorporate information from hand labeling or human prediction of the validation dataset or test data records.\nBecause this is a code competition, the public and private test set is unknown. Any attempt to identify and/or label the test plays would be considered against the spirit of the competition.
\nHope that helps. If you care to elaborate more on how you intend to use external data, I could clarify more specifically.
", "votes": 3, "replies": [] }, { "id": 2120256, "author_name": "Camaro", "author_url": "", "post_date": "2023-01-29T13:11:36.913000", "content": "Hi @robikscube,
\nHow about other Kaggle competitons' dataset, like DFL? It is allowed to use?
\nhttps://www.kaggle.com/competitions/dfl-bundesliga-data-shootout
@bamps53 - The DFL competition rules state that the dataset provided should only be used for the specific competition and should not be used for any other purpose. Because of those restrictions I would suggest you not use it in this competition.
\nhttps://www.kaggle.com/competitions/dfl-bundesliga-data-shootout/rules
\nSection 2.7.A: \nA. Data Access and Use. You may access and use the Competition Data only for participating in the Competition and on Kaggle.com forums...\n@robikscube The rules for older NFL competitions also have a similar statement to the DFL competition
\nhttps://www.kaggle.com/competitions/nfl-impact-detection/rules
\n\n\nParticipants in the Data Science Theme may access and use the Competition Data only for participating in the Data Science Theme and on Kaggle.com forums. The Sponsor and Named Partners reserve the right to disqualify any such Participant who uses the Competition Data other than as permitted by the Data Science Theme and these Official Rules. For the avoidance of doubt, Competition Data may include video footage from certain NFL games, and all such footage may be accessed and used only for participating in the Data Science Theme.
\n
Based on my understanding of the rules and your comment above, data from older NFL competitions cannot be used for training our models. Could you please confirm this?
", "votes": 1, "replies": [] }, { "id": 2133582, "author_name": "Rob Mulla", "author_url": "", "post_date": "2023-02-07T13:49:24.600000", "content": "Good catch. Considering this competition is a continuation of the NFL's \"First and Future\" efforts, the use of the prior \"First and Future\" competition data for this competition is ok.
", "votes": 1, "replies": [] } ] } ] }, { "id": 2127371, "author_name": "Zachary Orr", "author_url": "", "post_date": "2023-02-02T20:53:35.617000", "content": "How Exciting!
", "votes": 0, "replies": [] }, { "id": 2059094, "author_name": "pineapple", "author_url": "", "post_date": "2022-12-08T13:30:55.607000", "content": "so exciting go eaglesss
", "votes": 2, "replies": [] }, { "id": 2058629, "author_name": "Simon Alerdic", "author_url": "", "post_date": "2022-12-08T04:59:04.363000", "content": "@robikscube
\nI've a question about the price money for 1st place to 5th place. I think it is unfair to offer the 1st place so much (50%) credit.
1st Place - $50,000\n2nd Place - $25,000\n3rd Place - $13,000\n4th Place - $7,000\n5th Place - $5,000\nAs we see that, normally, the scores between top places don't vary much, so there is no such that level of nobility for particular top solutions. I am cordially suggesting to distribute the price money among top 10 for this high budget competition.
", "votes": 0, "replies": [ { "id": 2059178, "author_name": "Rob Mulla", "author_url": "", "post_date": "2022-12-08T14:34:52.310000", "content": "Hey @simonalerdic - I apprecaite your feedback! You make an interesting point. We may consider it for future competitions. The prize structure for this competition is set and we will not be changing it. Good luck!
", "votes": 1, "replies": [] }, { "id": 2091253, "author_name": "Qishen Ha", "author_url": "", "post_date": "2023-01-08T07:12:30.370000", "content": "If you checked last year's NFL competition: https://www.kaggle.com/competitions/nfl-health-and-safety-helmet-assignment/leaderboard
\nand last last year's NFL competition: https://www.kaggle.com/competitions/nfl-impact-detection/leaderboard
\nyou wouldn't post this…
let us do the best we can in this competition as the intention has always been to test our skills and continuous improvement!
", "votes": 0, "replies": [] } ] }, { "id": 2154460, "author_name": "HaijiangY", "author_url": "", "post_date": "2023-02-22T03:04:53.207000", "content": "How Exciting!
", "votes": -1, "replies": [] }, { "id": 2718145, "author_name": "trailblazer010", "author_url": "", "post_date": "2024-03-27T01:57:45.730000", "content": "Hello, I participated in kaggle's 1st and Future - Player Contact Detection competition and achieved top8%, a bronze medal score of 66/939. Now I need a certificate to prove my score. Can you tell me where I can get it? Have you obtained this certification? Thank you very much.
", "votes": 0, "replies": [] }, { "id": 2165124, "author_name": "ThreeLie", "author_url": "", "post_date": "2023-03-02T02:21:22.917000", "content": "Exciting and interesting!
", "votes": 0, "replies": [] }, { "id": 2159345, "author_name": "Niko NN", "author_url": "", "post_date": "2023-02-25T17:34:41.937000", "content": "Great competition!
", "votes": 0, "replies": [] }, { "id": 2152476, "author_name": "Chen Lin", "author_url": "", "post_date": "2023-02-20T20:15:40.363000", "content": "May I ask where the sensor that records tracking data is located on the player? Is it in the helmet? Or is it on the chest? Or are the different data coming from multiple sensors on multiple parts of the body?
", "votes": 0, "replies": [ { "id": 2152519, "author_name": "chris", "author_url": "", "post_date": "2023-02-20T20:58:47.543000", "content": "I believe the sensors are in their shoulder pads, but that may be old information (I see articles on google from 2015-2020 about it, but no newer ones)
", "votes": 1, "replies": [ { "id": 2152612, "author_name": "Chen Lin", "author_url": "", "post_date": "2023-02-20T22:47:20.620000", "content": "Do you know then approximately what the furthest distance between the two shoulders is if two players are in foot to foot contact? I'm not sure how the height should be converted to the distance units here.
", "votes": 0, "replies": [] } ] }, { "id": 2153972, "author_name": "Rob Mulla", "author_url": "", "post_date": "2023-02-21T18:36:51.300000", "content": "@chenlin1999 - To confirm the answer to your question about sensor locations, you can refer to this: https://operations.nfl.com/gameday/technology/nfl-next-gen-stats/
\n\"2–3 radio-frequency identification (RFID) tags installed into the players’ shoulder pads\"
Thank you!
", "votes": 0, "replies": [] } ] } ] }, { "id": 2151281, "author_name": "Yan Zhang", "author_url": "", "post_date": "2023-02-20T01:44:16.840000", "content": "I just realized there are some helmets missed for both sideline and endzone views (may because they are too small in the sideline view) that are postive contacts in the label csv. For example, game_play = '58173_003606' and frame = 710-716. Will the test data have the similar cases?
", "votes": 0, "replies": [ { "id": 2156898, "author_name": "Rob Mulla", "author_url": "", "post_date": "2023-02-23T15:50:18.457000", "content": "@annealer - yes, you can assume the test data is similar in nature to the training dataset.
", "votes": 0, "replies": [] } ] }, { "id": 2143939, "author_name": "Durgesh Chaubey", "author_url": "", "post_date": "2023-02-14T16:37:22.127000", "content": "its exciting.
", "votes": 0, "replies": [] }, { "id": 2135270, "author_name": "Haoran Zhao", "author_url": "", "post_date": "2023-02-08T14:35:01.870000", "content": "I'm new to kaggle and this project feels really interesting! How do I get started?？
", "votes": 0, "replies": [] }, { "id": 2092884, "author_name": "Byron J Ehrke", "author_url": "", "post_date": "2023-01-09T17:12:02.227000", "content": "Looks interesting to try!
", "votes": 0, "replies": [] }, { "id": 2090436, "author_name": "Geethu123", "author_url": "", "post_date": "2023-01-07T10:34:50.577000", "content": "Good one and very really interesting
", "votes": 0, "replies": [] }, { "id": 2150548, "author_name": "", "author_url": "", "post_date": "2023-02-19T11:08:10.103000", "content": "", "votes": 0, "replies": [] }, { "id": 2143835, "author_name": "", "author_url": "", "post_date": "2023-02-14T14:39:25.507000", "content": "", "votes": 0, "replies": [] }, { "id": 2068529, "author_name": "", "author_url": "", "post_date": "2022-12-18T04:54:12.177000", "content": "", "votes": 0, "replies": [ { "id": 2069367, "author_name": "Rob Mulla", "author_url": "", "post_date": "2022-12-18T21:50:16.470000", "content": "@palantirha - \"frame\" refers to a still image which makes up the video: https://en.wikipedia.org/wiki/Film_frame
\nHave you read the data description? You might find it helpful. https://www.kaggle.com/competitions/nfl-player-contact-detection/data
", "votes": 0, "replies": [ { "id": 2069906, "author_name": "", "author_url": "", "post_date": "2022-12-19T12:33:42.480000", "content": "", "votes": 1, "replies": [] } ] } ] }, { "id": 2062465, "author_name": "", "author_url": "", "post_date": "2022-12-12T05:56:35.570000", "content": "", "votes": 2, "replies": [] }, { "id": 2101875, "author_name": "", "author_url": "", "post_date": "2023-01-16T08:49:24.597000", "content": "", "votes": 1, "replies": [] }, { "id": 2061239, "author_name": "", "author_url": "", "post_date": "2022-12-10T21:27:49.140000", "content": "", "votes": 1, "replies": [] } ], "raw_markdown_by_id": { "2056202": "Welcome to the NFL Health and Safety - Player Contact Detection competition!\n\nYet again, we are looking forward to seeing how kagglers attack this problem!\n\nIf you haven't already, I encourage you to to read the [evaluation details](https://www.kaggle.com/competitions/nfl-player-contact-detection/overview/evaluation), [data description](https://www.kaggle.com/competitions/nfl-player-contact-detection/data) and our [starter guide notebook](https://www.kaggle.com/code/robikscube/nfl-player-contact-detection-getting-started).\n\nFeel free to post any questions you might have. @samhuddleston, @brittevans and I will try our best to respond promptly.\n\nThanks for joining and good luck!", "2059228": "@robikscube\nIn train_baseline_helmets.csv, I find video = `58226_003009_Endzone.mp4` and `58226_003009_Endzone2.mp4`. \nBut in train folder I can only find `58226_003009_Endzone.mp4`. What's their difference?", "2056271": "@robikscube Thanks for hosting this competition! Hope to see your baseline solution at the twitch live stream ", "2154353": "May I also ask, are the cameras' positions stationary and only pan/zoom during the game or are cameras moving as well?", "2128963": "@robikscube \nI'm not sure how timestamps are syncted between multiple data.\nIs this correct picture of timestamps?\n\n1. actual video started at `start_time` (frame=0)\n1. tracking is started at `snap_time` (frame=300)\n2. helmet labels are started at frame=290\n\n", "2124315": "where is the test.csv? I mean I build the model,but there is no test.csv.Is the testset data need to split from train dataset?", "2100666": "@robikscube \nI have concerns about the player's tracking dataset. The step feature has a max value of 692 which means those videos should have 69.2 seconds of duration! Of the videos, we have the longest one has 22.15 seconds duration. Why is this happening? Also, I think I understand this, but there are negative steps. I reason that they are before the game starts and from 0 starts the game. I would appreciate some explanation!", "2100361": "Hey Rob, can we get the baseline helmets for All29?", "2067191": "@robikscube\nThe dataset description page states that there is a column called \"event\" in [train/test]_player_tracking.csv, but the actual data does not contain it. Is this a typographical error in the description?", "2065288": "@robikscube \nRegarding \"Submission Limits\" in Rules section, is the current description correct? \n> There is no daily submission limit. Only the most recent submission by a team will be evaluated after the challenge deadline.", "2058271": "How Exciting!", "2056749": "It is great opportunity to update knowledge. Let's move.", "2132995": "I'm new on kaggle i don't know how to do ,here in data many csv files which one we have to use and can i start on google colab ?", "2115065": "@robikscube quick questions - not sure if it was already covered somewhere:\n\n- Is it allowed to use old NFL competition data?\n- Is it allowed to use Yolo models? I am a bit concerned given the open source requirement in this competition without commercial restriction, and the strict GPL 3 restrictions from Yolo - meaning we would also need to license this solution as GPL.", "2127371": "How Exciting!\n\n\n", "2059094": "so exciting go eaglesss", "2058629": "@robikscube \nI've a question about the price money for 1st place to 5th place. I think it is unfair to offer the 1st place so much (50%) credit. \n\n```\n1st Place - $50,000\n2nd Place - $25,000\n3rd Place - $13,000\n4th Place - $7,000\n5th Place - $5,000\n```\n\nAs we see that, normally, the scores between top places don't vary much, so there is no such **that level of nobility** for particular top solutions. I am cordially suggesting to distribute the price money among **top 10** for this high budget competition. ", "2154460": "How Exciting!", "2718145": "Hello, I participated in kaggle's 1st and Future - Player Contact Detection competition and achieved top8%, a bronze medal score of 66/939. Now I need a certificate to prove my score. Can you tell me where I can get it? Have you obtained this certification? Thank you very much.", "2165124": "Exciting and interesting!", "2159345": "Great competition!", "2152476": "May I ask where the sensor that records tracking data is located on the player? Is it in the helmet? Or is it on the chest? Or are the different data coming from multiple sensors on multiple parts of the body?", "2151281": "I just realized there are some helmets missed for both sideline and endzone views (may because they are too small in the sideline view) that are postive contacts in the label csv. For example, game_play = '58173_003606' and frame = 710-716. Will the test data have the similar cases?", "2143939": "its exciting.", "2135270": "I'm new to kaggle and this project feels really interesting! How do I get started?？", "2092884": "Looks interesting to try!", "2090436": "Good one and very really interesting ", "2150548": "", "2143835": "", "2068529": "", "2062465": "", "2101875": "Thank u for hosting", "2061239": "**Thanks a lot!! **@robikscube " } }