Hello my name is Hemankit! I am new to kaggle competitions but not to Data science. I joined this competition in the hopes of learning something new and putting my data science knowledge to work. I have a question about the model building. Do I need the model to make a prediction about whether a tomogram has bacteria with or without motors or do I need to have a model that checks directly the condition -1.0 for motor axis and once it declares that the bacteria has motors then I would use a localization model to find those coordinates? To put it simply am I trying to use a classification ML model and localization model together or is it more just working on the localization model once my algorithm checks the motor axis features of a test tomo_id data if the values are not -1.0 if it is then the localization model does nothing. Please Clarify. Thank You!
", "messages": [ { "id": "3157161", "postDate": "03/23/2025 03:57:47", "content": "Hello my name is Hemankit! I am new to kaggle competitions but not to Data science. I joined this competition in the hopes of learning something new and putting my data science knowledge to work. I have a question about the model building. Do I need the model to make a prediction about whether a tomogram has bacteria with or without motors or do I need to have a model that checks directly the condition -1.0 for motor axis and once it declares that the bacteria has motors then I would use a localization model to find those coordinates? To put it simply am I trying to use a classification ML model and localization model together or is it more just working on the localization model once my algorithm checks the motor axis features of a test tomo_id data if the values are not -1.0 if it is then the localization model does nothing. Please Clarify. Thank You!
", "rawMarkdown": "Hello my name is Hemankit! I am new to kaggle competitions but not to Data science. I joined this competition in the hopes of learning something new and putting my data science knowledge to work. I have a question about the model building. Do I need the model to make a prediction about whether a tomogram has bacteria with or without motors or do I need to have a model that checks directly the condition -1.0 for motor axis and once it declares that the bacteria has motors then I would use a localization model to find those coordinates? To put it simply am I trying to use a classification ML model and localization model together or is it more just working on the localization model once my algorithm checks the motor axis features of a test tomo_id data if the values are not -1.0 if it is then the localization model does nothing. Please Clarify. Thank You!", "votes": null }, { "id": "3157426", "postDate": "03/23/2025 12:30:00", "content": "scientists discovered that there are motors that exist in bacteria, the idea of the competition is just to detect the location of the motors in a tomogram, if the model didn't detect any motor in the tomogram you should return these coordinates (-1,-1,-1)
\nwhat is a tomogram?
\nAns) if you want it in simple imagine that there is a 3d freezed block that contain bacteria and other structures, we cut this block into multiple slices (around 500 slices for example, it differs from tomogram to other ) and take a 2d picture for each slice, these slices combined are a tomogram
\nso you need to detect the x,y location of the motor, z axis represent the number of slice in a tomogram
\nso you need to focus on an object detection approach for motors and you don't have to detect or look on bacteria
", "rawMarkdown": "scientists discovered that there are motors that exist in bacteria, the idea of the competition is just to detect the location of the motors in a tomogram, if the model didn't detect any motor in the tomogram you should return these coordinates (-1,-1,-1)\n\nwhat is a tomogram?\n\nAns) if you want it in simple imagine that there is a 3d freezed block that contain bacteria and other structures, we cut this block into multiple slices (around 500 slices for example, it differs from tomogram to other ) and take a 2d picture for each slice, these slices combined are a tomogram\n\nso you need to detect the x,y location of the motor, z axis represent the number of slice in a tomogram\n\nso you need to focus on an object detection approach for motors and you don't have to detect or look on bacteria", "votes": null }, { "id": "3157673", "postDate": "03/23/2025 17:19:48", "content": "The test dataset doesn't have the information about motor presence, it's just tomogram slices. So your model needs to predict motor presence and the location when it detects a motor. You can check the test and sample_submission data in the competition dataset to see how the model is supposed to make predictions.
", "rawMarkdown": "The test dataset doesn't have the information about motor presence, it's just tomogram slices. So your model needs to predict motor presence and the location when it detects a motor. You can check the test and sample_submission data in the competition dataset to see how the model is supposed to make predictions.", "votes": null }, { "id": "3158508", "postDate": "03/24/2025 15:57:14", "content": "Oh, I see. that condition is implied based on the results of the classification. My idea was to use a classification algorithm to first classify whether the bacteria have motors or not and then transition to a localization model that will find the coordinates. So, I guess based on what you told me, and please correct me if I am wrong, if the algorithm classified it as having no motors it should output -1 for all three of those features (motor axis x, y, and z) for a new tomogram. in that case it would not make much sense to use the motor axis features from the training set to classify bacteria having motors or not? I would need to utilize the other features for it right? I just want to clarify this so that I know how to proceed. I hope this is not asking too much:)
", "rawMarkdown": "Oh, I see. that condition is implied based on the results of the classification. My idea was to use a classification algorithm to first classify whether the bacteria have motors or not and then transition to a localization model that will find the coordinates. So, I guess based on what you told me, and please correct me if I am wrong, if the algorithm classified it as having no motors it should output -1 for all three of those features (motor axis x, y, and z) for a new tomogram. in that case it would not make much sense to use the motor axis features from the training set to classify bacteria having motors or not? I would need to utilize the other features for it right? I just want to clarify this so that I know how to proceed. I hope this is not asking too much:)", "votes": null }, { "id": "3158602", "postDate": "03/24/2025 17:32:54", "content": "the x,y,z coordinates are the target not a feature so yeah you should not train on them
", "rawMarkdown": "the x,y,z coordinates are the target not a feature so yeah you should not train on them", "votes": null } ], "comments": [ { "id": 3157426, "author_name": "tahaalshatiri", "author_url": "", "post_date": "03/23/2025 12:30:00", "content": "scientists discovered that there are motors that exist in bacteria, the idea of the competition is just to detect the location of the motors in a tomogram, if the model didn't detect any motor in the tomogram you should return these coordinates (-1,-1,-1)
\nwhat is a tomogram?
\nAns) if you want it in simple imagine that there is a 3d freezed block that contain bacteria and other structures, we cut this block into multiple slices (around 500 slices for example, it differs from tomogram to other ) and take a 2d picture for each slice, these slices combined are a tomogram
\nso you need to detect the x,y location of the motor, z axis represent the number of slice in a tomogram
\nso you need to focus on an object detection approach for motors and you don't have to detect or look on bacteria
", "votes": null, "replies": [ { "id": 3158508, "author_name": "hemankit", "author_url": "", "post_date": "03/24/2025 15:57:14", "content": "Oh, I see. that condition is implied based on the results of the classification. My idea was to use a classification algorithm to first classify whether the bacteria have motors or not and then transition to a localization model that will find the coordinates. So, I guess based on what you told me, and please correct me if I am wrong, if the algorithm classified it as having no motors it should output -1 for all three of those features (motor axis x, y, and z) for a new tomogram. in that case it would not make much sense to use the motor axis features from the training set to classify bacteria having motors or not? I would need to utilize the other features for it right? I just want to clarify this so that I know how to proceed. I hope this is not asking too much:)
", "votes": null, "replies": [ { "id": 3158602, "author_name": "tahaalshatiri", "author_url": "", "post_date": "03/24/2025 17:32:54", "content": "the x,y,z coordinates are the target not a feature so yeah you should not train on them
", "votes": null, "replies": [] } ] } ] }, { "id": 3157673, "author_name": "tennogh", "author_url": "", "post_date": "03/23/2025 17:19:48", "content": "The test dataset doesn't have the information about motor presence, it's just tomogram slices. So your model needs to predict motor presence and the location when it detects a motor. You can check the test and sample_submission data in the competition dataset to see how the model is supposed to make predictions.
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "3157161": "Hello my name is Hemankit! I am new to kaggle competitions but not to Data science. I joined this competition in the hopes of learning something new and putting my data science knowledge to work. I have a question about the model building. Do I need the model to make a prediction about whether a tomogram has bacteria with or without motors or do I need to have a model that checks directly the condition -1.0 for motor axis and once it declares that the bacteria has motors then I would use a localization model to find those coordinates? To put it simply am I trying to use a classification ML model and localization model together or is it more just working on the localization model once my algorithm checks the motor axis features of a test tomo_id data if the values are not -1.0 if it is then the localization model does nothing. Please Clarify. Thank You!", "3157426": "scientists discovered that there are motors that exist in bacteria, the idea of the competition is just to detect the location of the motors in a tomogram, if the model didn't detect any motor in the tomogram you should return these coordinates (-1,-1,-1)\n\nwhat is a tomogram?\n\nAns) if you want it in simple imagine that there is a 3d freezed block that contain bacteria and other structures, we cut this block into multiple slices (around 500 slices for example, it differs from tomogram to other ) and take a 2d picture for each slice, these slices combined are a tomogram\n\nso you need to detect the x,y location of the motor, z axis represent the number of slice in a tomogram\n\nso you need to focus on an object detection approach for motors and you don't have to detect or look on bacteria", "3157673": "The test dataset doesn't have the information about motor presence, it's just tomogram slices. So your model needs to predict motor presence and the location when it detects a motor. You can check the test and sample_submission data in the competition dataset to see how the model is supposed to make predictions.", "3158508": "Oh, I see. that condition is implied based on the results of the classification. My idea was to use a classification algorithm to first classify whether the bacteria have motors or not and then transition to a localization model that will find the coordinates. So, I guess based on what you told me, and please correct me if I am wrong, if the algorithm classified it as having no motors it should output -1 for all three of those features (motor axis x, y, and z) for a new tomogram. in that case it would not make much sense to use the motor axis features from the training set to classify bacteria having motors or not? I would need to utilize the other features for it right? I just want to clarify this so that I know how to proceed. I hope this is not asking too much:)", "3158602": "the x,y,z coordinates are the target not a feature so yeah you should not train on them" }, "source": "meta" }