![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F846065%2F9316b6942a9d2b00b855447fcfb2db3c%2Ftab1.png?generation=1588966674153611&alt=media\")
\n![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F846065%2F074af3b57cd3ba35bd21b6faf9d7fde0%2Ftab2.png?generation=1588966682437024&alt=media\")
I did not realize there is an extensive description of the dataset provided in a pre-print. It includes a lot of useful information that we can hopefully use. For example, take a look at the following two tables:\n\n
Have ideas on how we can make use of this data? Comment below! I'd love to hear your ideas
\n\nThe full preprint is here https://arxiv.org/pdf/2005.02162.pdf
\n\nCitation:\nDavid, E., et al. \"Global Wheat Head Detection (GWHD) dataset: a large and diverse dataset of high resolution RGB labelled images to develop and benchmark wheat head detection methods.\" arXiv preprint arXiv:2005.02162 (2020).
", "messages": [ { "id": 838781, "postDate": "2020-05-08T19:38:16.953Z", "content": "I did not realize there is an extensive description of the dataset provided in a pre-print. It includes a lot of useful information that we can hopefully use. For example, take a look at the following two tables:\n\n
Have ideas on how we can make use of this data? Comment below! I'd love to hear your ideas
\n\nThe full preprint is here https://arxiv.org/pdf/2005.02162.pdf
\n\nCitation:\nDavid, E., et al. \"Global Wheat Head Detection (GWHD) dataset: a large and diverse dataset of high resolution RGB labelled images to develop and benchmark wheat head detection methods.\" arXiv preprint arXiv:2005.02162 (2020).
", "rawMarkdown": "I did not realize there is an extensive description of the dataset provided in a pre-print. It includes a lot of useful information that we can hopefully use. For example, take a look at the following two tables:\n\n\n\n\nHave ideas on how we can make use of this data? Comment below! I'd love to hear your ideas\n\nThe full preprint is here https://arxiv.org/pdf/2005.02162.pdf\n\nCitation:\nDavid, E., et al. \"Global Wheat Head Detection (GWHD) dataset: a large and diverse dataset of high resolution RGB labelled images to develop and benchmark wheat head detection methods.\" arXiv preprint arXiv:2005.02162 (2020).\n", "votes": 8 }, { "id": 839474, "postDate": "2020-05-09T12:08:25.903Z", "content": "Won't using such location specific features negatively affect a model's generalizability in some way (assuming there is a way of using this meta data for training)?
\n\n\n\n\nUsing worldwide data, you will focus on a generalized solution to estimate the number and size of wheat heads. To better gauge the performance for unseen genotypes, environments, and observational conditions, the training dataset covers multiple regions.
\n
I personally don't think it'd be a good idea to use this meta given the above. What do you think?
", "rawMarkdown": "Won't using such location specific features negatively affect a model's generalizability in some way (assuming there is a way of using this meta data for training)?\n> Using worldwide data, you will focus on a generalized solution to estimate the number and size of wheat heads. To better gauge the performance for unseen genotypes, environments, and observational conditions, the training dataset covers multiple regions.\n\nI personally don't think it'd be a good idea to use this meta given the above. What do you think?", "votes": 2, "replies": [ { "id": 840138, "postDate": "2020-05-09T19:40:55.750Z", "content": "The thing is, we know exactly which \"sub-datasets\" are provided in the training and test sets. Therefor, we can actually use this information to help in our generalization.
\n\nFor example, maybe someone could take satellite imagery from each of the locations to see and understand the differences in the appearance of each site. Another example is that since many different cameras are used, maybe you can transform the images based on their focal length, FOV, etc. to standardize the images and make them appear more similar.
\n\nThese are just my initial thoughts, so there's probably lots of things I have not accounted for.
", "rawMarkdown": "The thing is, we know exactly which \"sub-datasets\" are provided in the training and test sets. Therefor, we can actually use this information to help in our generalization.\n\nFor example, maybe someone could take satellite imagery from each of the locations to see and understand the differences in the appearance of each site. Another example is that since many different cameras are used, maybe you can transform the images based on their focal length, FOV, etc. to standardize the images and make them appear more similar.\n\nThese are just my initial thoughts, so there's probably lots of things I have not accounted for.", "votes": 2 }, { "id": 840678, "postDate": "2020-05-10T09:09:15.553Z", "rawMarkdown": "", "isDeleted": true } ] }, { "id": 840108, "postDate": "2020-05-09T19:16:39.077Z", "content": "I think you may need to use some of the data, i am quite new to machine learning but have seen some comments about calibration of images when using different camera types, not sure if this is of any help
", "rawMarkdown": "I think you may need to use some of the data, i am quite new to machine learning but have seen some comments about calibration of images when using different camera types, not sure if this is of any help" }, { "id": 838872, "postDate": "2020-05-08T21:14:56.860Z", "content": "Nice :)
", "rawMarkdown": "Nice :)" } ], "comments": [ { "id": 839474, "author_name": "ms", "author_url": "", "post_date": "2020-05-09T12:08:25.903000", "content": "Won't using such location specific features negatively affect a model's generalizability in some way (assuming there is a way of using this meta data for training)?
\n\n\n\n\nUsing worldwide data, you will focus on a generalized solution to estimate the number and size of wheat heads. To better gauge the performance for unseen genotypes, environments, and observational conditions, the training dataset covers multiple regions.
\n
I personally don't think it'd be a good idea to use this meta given the above. What do you think?
", "votes": 2, "replies": [ { "id": 840138, "author_name": "Matt", "author_url": "", "post_date": "2020-05-09T19:40:55.750000", "content": "The thing is, we know exactly which \"sub-datasets\" are provided in the training and test sets. Therefor, we can actually use this information to help in our generalization.
\n\nFor example, maybe someone could take satellite imagery from each of the locations to see and understand the differences in the appearance of each site. Another example is that since many different cameras are used, maybe you can transform the images based on their focal length, FOV, etc. to standardize the images and make them appear more similar.
\n\nThese are just my initial thoughts, so there's probably lots of things I have not accounted for.
", "votes": 2, "replies": [] }, { "id": 840678, "author_name": "", "author_url": "", "post_date": "2020-05-10T09:09:15.553000", "content": "", "votes": 0, "replies": [] } ] }, { "id": 840108, "author_name": "John Scanlan", "author_url": "", "post_date": "2020-05-09T19:16:39.077000", "content": "I think you may need to use some of the data, i am quite new to machine learning but have seen some comments about calibration of images when using different camera types, not sure if this is of any help
", "votes": 0, "replies": [] }, { "id": 838872, "author_name": "Alberto Maria Falletta", "author_url": "", "post_date": "2020-05-08T21:14:56.860000", "content": "Nice :)
", "votes": 0, "replies": [] } ], "raw_markdown_by_id": { "838781": "I did not realize there is an extensive description of the dataset provided in a pre-print. It includes a lot of useful information that we can hopefully use. For example, take a look at the following two tables:\n\n\n\n\nHave ideas on how we can make use of this data? Comment below! I'd love to hear your ideas\n\nThe full preprint is here https://arxiv.org/pdf/2005.02162.pdf\n\nCitation:\nDavid, E., et al. \"Global Wheat Head Detection (GWHD) dataset: a large and diverse dataset of high resolution RGB labelled images to develop and benchmark wheat head detection methods.\" arXiv preprint arXiv:2005.02162 (2020).\n", "839474": "Won't using such location specific features negatively affect a model's generalizability in some way (assuming there is a way of using this meta data for training)?\n> Using worldwide data, you will focus on a generalized solution to estimate the number and size of wheat heads. To better gauge the performance for unseen genotypes, environments, and observational conditions, the training dataset covers multiple regions.\n\nI personally don't think it'd be a good idea to use this meta given the above. What do you think?", "840108": "I think you may need to use some of the data, i am quite new to machine learning but have seen some comments about calibration of images when using different camera types, not sure if this is of any help", "838872": "Nice :)" } }