![\"enter](\"https://i.imgur.com/OUCbJCF.png\" "\\"\\"")
Welcome to Cdiscount's Image Classification Challenge! Both CDiscount and Kaggle are excited for to host this competition, which provides a challenging data set of over 15 million images representing 9 million products in over 5,000 categories.
\n\nThis challenge provides the opportunity to work with the BSON file format. Please check out the example kernel, which demonstrates how to read and process the file iteratively. You may also consider getting started on the train_example.bson file, which includes the first 100 records in the training file.
\n\nPlease feel free to ask your questions in this thread.
\n\nGood luck!
", "messages": [ { "id": 221285, "postDate": "2017-09-14T17:21:43.747Z", "content": "Welcome to Cdiscount's Image Classification Challenge! Both CDiscount and Kaggle are excited for to host this competition, which provides a challenging data set of over 15 million images representing 9 million products in over 5,000 categories.
\n\nThis challenge provides the opportunity to work with the BSON file format. Please check out the example kernel, which demonstrates how to read and process the file iteratively. You may also consider getting started on the train_example.bson file, which includes the first 100 records in the training file.
\n\nPlease feel free to ask your questions in this thread.
\n\nGood luck!
", "rawMarkdown": "Welcome to Cdiscount's Image Classification Challenge! Both CDiscount and Kaggle are excited for to host this competition, which provides a challenging data set of over 15 million images representing 9 million products in over 5,000 categories.\n\nThis challenge provides the opportunity to work with the BSON file format. Please check out the [example kernel,][1] which demonstrates how to read and process the file iteratively. You may also consider getting started on the train_example.bson file, which includes the first 100 records in the training file.\n\nPlease feel free to ask your questions in this thread.\n\nGood luck!\n\n\n [1]: https://www.kaggle.com/inversion/processing-bson-files", "votes": 11 }, { "id": 225224, "postDate": "2017-09-28T14:52:41.500Z", "content": "Hi @inversion! Can you say a few words about your similarity benchmark? It's impressive!
", "rawMarkdown": "Hi @inversion! Can you say a few words about your similarity benchmark? It's impressive!", "votes": 1, "replies": [ { "id": 225384, "postDate": "2017-09-28T21:38:41.727Z", "content": "Thanks!! It's mainly brute force. I used a perceptual hashing algorithm to generate a signature for each image, and then predicted categories based on signature distances. It didn't require a GPU, but took ~10 hours on a 64-core CPU.
", "rawMarkdown": "Thanks!! It's mainly brute force. I used a [perceptual hashing][1] algorithm to generate a signature for each image, and then predicted categories based on signature distances. It didn't require a GPU, but took ~10 hours on a 64-core CPU.\n\n [1]: https://en.wikipedia.org/wiki/Perceptual_hashing", "votes": 6 }, { "id": 225425, "postDate": "2017-09-29T02:40:31.353Z", "content": "Thanks!
", "rawMarkdown": "Thanks!" } ] }, { "id": 242689, "postDate": "2017-11-12T11:40:25.667Z", "content": "I do not have 60GB+ of space to use in my pc. I do have GPUs and RAM to use though, any ideas on how to get access to the data?
", "rawMarkdown": "I do not have 60GB+ of space to use in my pc. I do have GPUs and RAM to use though, any ideas on how to get access to the data?" }, { "id": 240232, "postDate": "2017-11-06T01:47:20.507Z", "content": "Hi, how can I access the data from Google Cloud?
", "rawMarkdown": "Hi, how can I access the data from Google Cloud? ", "replies": [ { "id": 243807, "postDate": "2017-11-14T23:12:48.853Z", "content": "Hi Bella - Sorry for the slow response. I was traveling last week. We don't have the data loaded in Google, other than in the Kernels environment. So, you can either access it through kernels, or create your own google cloud environment and transfer the data over.
", "rawMarkdown": "Hi Bella - Sorry for the slow response. I was traveling last week. We don't have the data loaded in Google, other than in the Kernels environment. So, you can either access it through kernels, or create your own google cloud environment and transfer the data over." } ] }, { "id": 226883, "postDate": "2017-10-03T08:51:12.130Z", "content": "Hi,\nI am looking for the hash of the files.
", "rawMarkdown": "Hi,\nI am looking for the hash of the files. ", "replies": [ { "id": 226884, "postDate": "2017-10-03T08:54:07.737Z", "content": "I found the answer here: https://www.kaggle.com/c/cdiscount-image-classification-challenge/discussion/40052
", "rawMarkdown": "I found the answer here: https://www.kaggle.com/c/cdiscount-image-classification-challenge/discussion/40052" } ] }, { "id": 223238, "postDate": "2017-09-21T14:51:32.693Z", "content": "Could I get started with only one single NVIDIA 1080TI GPU？
", "rawMarkdown": "Could I get started with only one single NVIDIA 1080TI GPU？", "replies": [ { "id": 223362, "postDate": "2017-09-22T03:01:13.420Z", "content": "Did you get started? I also only have one single 1080TI gpu.
", "rawMarkdown": "Did you get started? I also only have one single 1080TI gpu." } ] }, { "id": 222713, "postDate": "2017-09-19T18:50:36.643Z", "content": "Hi! The number of training examples (images) per class (category) varies a lot in the training set, from over 80000 in the classes with most training examples to just 12 in the class with the least. Can we assume that the testing set has a similar distribution or can a class with just a few images in the training set have thousands of images in the test set?
", "rawMarkdown": "Hi! The number of training examples (images) per class (category) varies a lot in the training set, from over 80000 in the classes with most training examples to just 12 in the class with the least. Can we assume that the testing set has a similar distribution or can a class with just a few images in the training set have thousands of images in the test set?", "replies": [ { "id": 222731, "postDate": "2017-09-19T20:00:18.427Z", "content": "Then only information we provide on this is that the the category_ids that are present in Private Test split are also all present in the Public Test split. No other assumptions have been provided.
", "rawMarkdown": "Then only information we provide on this is that the the category_ids that are present in Private Test split are also all present in the Public Test split. No other assumptions have been provided." } ] }, { "id": 222243, "postDate": "2017-09-18T06:39:23.693Z", "content": "Welcome to kaggle.
", "rawMarkdown": "Welcome to kaggle." }, { "id": 221415, "postDate": "2017-09-15T07:19:19.143Z", "rawMarkdown": "", "isDeleted": true, "replies": [ { "id": 221524, "postDate": "2017-09-15T15:17:56Z", "content": "Once you save a submission file from the kernel, you'll see the option to submit to the leader board.
\n\n
Please let me know if you run into any issues and I'll get that information to our engineering team. Thx!
", "rawMarkdown": "Please let me know if you run into any issues and I'll get that information to our engineering team. Thx!" }, { "id": 221542, "postDate": "2017-09-15T15:47:05.900Z", "rawMarkdown": "", "isDeleted": true }, { "id": 221552, "postDate": "2017-09-15T16:05:26.653Z", "content": "I need to verify if the functionality works with zipped files in kernels. In the mean time, can you try unzipped csv? Thx!
", "rawMarkdown": "I need to verify if the functionality works with zipped files in kernels. In the mean time, can you try unzipped csv? Thx!" }, { "id": 221556, "postDate": "2017-09-15T16:21:54.013Z", "rawMarkdown": "", "votes": 1, "isDeleted": true }, { "id": 222003, "postDate": "2017-09-17T07:32:44.253Z", "rawMarkdown": "", "isDeleted": true }, { "id": 222082, "postDate": "2017-09-17T12:32:49.337Z", "content": "Thanks. We're looking into this.
", "rawMarkdown": "Thanks. We're looking into this." } ] } ], "comments": [ { "id": 225224, "author_name": "Aloisio Dourado", "author_url": "", "post_date": "2017-09-28T14:52:41.500000", "content": "Hi @inversion! Can you say a few words about your similarity benchmark? It's impressive!
", "votes": 1, "replies": [ { "id": 225384, "author_name": "inversion", "author_url": "", "post_date": "2017-09-28T21:38:41.727000", "content": "Thanks!! It's mainly brute force. I used a perceptual hashing algorithm to generate a signature for each image, and then predicted categories based on signature distances. It didn't require a GPU, but took ~10 hours on a 64-core CPU.
", "votes": 6, "replies": [] }, { "id": 225425, "author_name": "Aloisio Dourado", "author_url": "", "post_date": "2017-09-29T02:40:31.353000", "content": "Thanks!
", "votes": 0, "replies": [] } ] }, { "id": 242689, "author_name": "Skinish", "author_url": "", "post_date": "2017-11-12T11:40:25.667000", "content": "I do not have 60GB+ of space to use in my pc. I do have GPUs and RAM to use though, any ideas on how to get access to the data?
", "votes": 0, "replies": [] }, { "id": 240232, "author_name": "Bella Wang", "author_url": "", "post_date": "2017-11-06T01:47:20.507000", "content": "Hi, how can I access the data from Google Cloud?
", "votes": 0, "replies": [ { "id": 243807, "author_name": "inversion", "author_url": "", "post_date": "2017-11-14T23:12:48.853000", "content": "Hi Bella - Sorry for the slow response. I was traveling last week. We don't have the data loaded in Google, other than in the Kernels environment. So, you can either access it through kernels, or create your own google cloud environment and transfer the data over.
", "votes": 0, "replies": [] } ] }, { "id": 226883, "author_name": "Beri", "author_url": "", "post_date": "2017-10-03T08:51:12.130000", "content": "Hi,\nI am looking for the hash of the files.
", "votes": 0, "replies": [ { "id": 226884, "author_name": "Beri", "author_url": "", "post_date": "2017-10-03T08:54:07.737000", "content": "I found the answer here: https://www.kaggle.com/c/cdiscount-image-classification-challenge/discussion/40052
", "votes": 0, "replies": [] } ] }, { "id": 223238, "author_name": "Currylee", "author_url": "", "post_date": "2017-09-21T14:51:32.693000", "content": "Could I get started with only one single NVIDIA 1080TI GPU？
", "votes": 0, "replies": [ { "id": 223362, "author_name": "Will", "author_url": "", "post_date": "2017-09-22T03:01:13.420000", "content": "Did you get started? I also only have one single 1080TI gpu.
", "votes": 0, "replies": [] } ] }, { "id": 222713, "author_name": "Per Isacson", "author_url": "", "post_date": "2017-09-19T18:50:36.643000", "content": "Hi! The number of training examples (images) per class (category) varies a lot in the training set, from over 80000 in the classes with most training examples to just 12 in the class with the least. Can we assume that the testing set has a similar distribution or can a class with just a few images in the training set have thousands of images in the test set?
", "votes": 0, "replies": [ { "id": 222731, "author_name": "inversion", "author_url": "", "post_date": "2017-09-19T20:00:18.427000", "content": "Then only information we provide on this is that the the category_ids that are present in Private Test split are also all present in the Public Test split. No other assumptions have been provided.
", "votes": 0, "replies": [] } ] }, { "id": 222243, "author_name": "Luiz Gustavo Mori", "author_url": "", "post_date": "2017-09-18T06:39:23.693000", "content": "Welcome to kaggle.
", "votes": 0, "replies": [] }, { "id": 221415, "author_name": "", "author_url": "", "post_date": "2017-09-15T07:19:19.143000", "content": "", "votes": 0, "replies": [ { "id": 221524, "author_name": "inversion", "author_url": "", "post_date": "2017-09-15T15:17:56", "content": "Once you save a submission file from the kernel, you'll see the option to submit to the leader board.
\n\n
Please let me know if you run into any issues and I'll get that information to our engineering team. Thx!
", "votes": 0, "replies": [] }, { "id": 221542, "author_name": "", "author_url": "", "post_date": "2017-09-15T15:47:05.900000", "content": "", "votes": 0, "replies": [] }, { "id": 221552, "author_name": "inversion", "author_url": "", "post_date": "2017-09-15T16:05:26.653000", "content": "I need to verify if the functionality works with zipped files in kernels. In the mean time, can you try unzipped csv? Thx!
", "votes": 0, "replies": [] }, { "id": 221556, "author_name": "", "author_url": "", "post_date": "2017-09-15T16:21:54.013000", "content": "", "votes": 1, "replies": [] }, { "id": 222003, "author_name": "", "author_url": "", "post_date": "2017-09-17T07:32:44.253000", "content": "", "votes": 0, "replies": [] }, { "id": 222082, "author_name": "inversion", "author_url": "", "post_date": "2017-09-17T12:32:49.337000", "content": "Thanks. We're looking into this.
", "votes": 0, "replies": [] } ] } ], "raw_markdown_by_id": { "221285": "Welcome to Cdiscount's Image Classification Challenge! Both CDiscount and Kaggle are excited for to host this competition, which provides a challenging data set of over 15 million images representing 9 million products in over 5,000 categories.\n\nThis challenge provides the opportunity to work with the BSON file format. Please check out the [example kernel,][1] which demonstrates how to read and process the file iteratively. You may also consider getting started on the train_example.bson file, which includes the first 100 records in the training file.\n\nPlease feel free to ask your questions in this thread.\n\nGood luck!\n\n\n [1]: https://www.kaggle.com/inversion/processing-bson-files", "225224": "Hi @inversion! Can you say a few words about your similarity benchmark? It's impressive!", "242689": "I do not have 60GB+ of space to use in my pc. I do have GPUs and RAM to use though, any ideas on how to get access to the data?", "240232": "Hi, how can I access the data from Google Cloud? ", "226883": "Hi,\nI am looking for the hash of the files. ", "223238": "Could I get started with only one single NVIDIA 1080TI GPU？", "222713": "Hi! The number of training examples (images) per class (category) varies a lot in the training set, from over 80000 in the classes with most training examples to just 12 in the class with the least. Can we assume that the testing set has a similar distribution or can a class with just a few images in the training set have thousands of images in the test set?", "222243": "Welcome to kaggle.", "221415": "" } }