{
  "total": 157,
  "topics": [
    {
      "id": 37137,
      "title": "Near-perfect Dice scores are okay",
      "comment_count": 44,
      "views": 0,
      "votes": 41
    },
    {
      "id": 37074,
      "title": "Official external data and pre-trained model thread",
      "comment_count": 49,
      "views": 0,
      "votes": 12
    },
    {
      "id": 364523,
      "title": "About Usage/License of Carvana dataset outside of competition",
      "comment_count": 0,
      "views": 0,
      "votes": 4
    },
    {
      "id": 37208,
      "title": "pytorch starter-kit (LB: 0.985) and share your experimental results here!",
      "comment_count": 165,
      "views": 0,
      "votes": 135
    },
    {
      "id": 274835,
      "title": "Kaggle carvana dataset usage and license",
      "comment_count": 0,
      "views": 0,
      "votes": 2
    },
    {
      "id": 47320,
      "title": "[Video] Gold medal solution by Sergey Mushinsky (Eng subtitles)",
      "comment_count": 1,
      "views": 0,
      "votes": 23
    },
    {
      "id": 194992,
      "title": "SGD vs RMSPROP vs ADAM vs OTHER OPTIMIZER on cifar10 dataset",
      "comment_count": 1,
      "views": 0,
      "votes": null
    },
    {
      "id": 38505,
      "title": "RMSProp vs SGD vs Adam vs ...",
      "comment_count": 22,
      "views": 0,
      "votes": 11
    },
    {
      "id": 39705,
      "title": "Submissions after merger deadline",
      "comment_count": 4,
      "views": 0,
      "votes": 0
    },
    {
      "id": 46176,
      "title": "1 st place solution - code",
      "comment_count": 1,
      "views": 0,
      "votes": 46
    },
    {
      "id": 40291,
      "title": "Trying to understand the \"smoothness\" in dice loss",
      "comment_count": 7,
      "views": 0,
      "votes": 1
    },
    {
      "id": 88240,
      "title": "My Kernel Need more hours to run",
      "comment_count": 0,
      "views": 0,
      "votes": 0
    },
    {
      "id": 37523,
      "title": "Keras starter (LB: 0.988)",
      "comment_count": 107,
      "views": 0,
      "votes": 102
    },
    {
      "id": 37415,
      "title": "Let's learn fp16 (half float) and multi-GPU in pytorch here!",
      "comment_count": 11,
      "views": 0,
      "votes": 14
    },
    {
      "id": 40199,
      "title": "3rd place solution (U-Net + Dilated Conv)",
      "comment_count": 25,
      "views": 0,
      "votes": 96
    },
    {
      "id": 37634,
      "title": "Want to speed up low GPU performance 1.5 times and don't have integrated video card? (no clickbait)",
      "comment_count": 6,
      "views": 0,
      "votes": 4
    },
    {
      "id": 54978,
      "title": "transforms the dataset into tfrecords that can be feed into deeplab model",
      "comment_count": 0,
      "views": 0,
      "votes": 0
    },
    {
      "id": 51686,
      "title": "Requesting permission to publish this competition data",
      "comment_count": 0,
      "views": 0,
      "votes": 2
    },
    {
      "id": 47872,
      "title": "Meetup in Mountain View, Jan 24",
      "comment_count": 0,
      "views": 0,
      "votes": 5
    },
    {
      "id": 39455,
      "title": "Dense CRF",
      "comment_count": 6,
      "views": 0,
      "votes": 2
    },
    {
      "id": 38207,
      "title": "\"all CUDA-capable devices are busy or unavailable\" when use heng's script",
      "comment_count": 5,
      "views": 0,
      "votes": 1
    },
    {
      "id": 39204,
      "title": "How to identify the \"difficult\" test images",
      "comment_count": 16,
      "views": 0,
      "votes": 22
    },
    {
      "id": 40121,
      "title": "comparing ranking and GPU",
      "comment_count": 77,
      "views": 0,
      "votes": 29
    },
    {
      "id": 40374,
      "title": "I keep getting evaluation exception when making late submissions",
      "comment_count": 7,
      "views": 0,
      "votes": null
    },
    {
      "id": 40241,
      "title": "Comparing ranking and framework",
      "comment_count": 14,
      "views": 0,
      "votes": 5
    },
    {
      "id": 39970,
      "title": "Evaluation Exception: Stream was not readable",
      "comment_count": 1,
      "views": 0,
      "votes": null
    },
    {
      "id": 40119,
      "title": "Overview of JbestDeepGooseFlops solution (3rd public)",
      "comment_count": 19,
      "views": 0,
      "votes": 34
    },
    {
      "id": 39951,
      "title": "weighing difficult pixel by focal loss",
      "comment_count": 6,
      "views": 0,
      "votes": 16
    },
    {
      "id": 40126,
      "title": "LB 11th(0.9971) solution overview ",
      "comment_count": 13,
      "views": 0,
      "votes": 33
    },
    {
      "id": 40118,
      "title": "Lessons  learned",
      "comment_count": 9,
      "views": 0,
      "votes": 32
    },
    {
      "id": 40144,
      "title": "LB0.9972 - 5th Place Overview",
      "comment_count": 5,
      "views": 0,
      "votes": 28
    },
    {
      "id": 40146,
      "title": "LB 15th (0.9970) Some new ideas (and a bug!)",
      "comment_count": 7,
      "views": 0,
      "votes": 25
    },
    {
      "id": 40523,
      "title": "In case your GPU is too cold",
      "comment_count": 0,
      "views": 0,
      "votes": 2
    },
    {
      "id": 40107,
      "title": "0.9970 Solution",
      "comment_count": 8,
      "views": 0,
      "votes": 35
    },
    {
      "id": 40123,
      "title": "0.9969 on a single GTX 1080",
      "comment_count": 11,
      "views": 0,
      "votes": 24
    },
    {
      "id": 40133,
      "title": "LB 10th place 0.9971 solution",
      "comment_count": 7,
      "views": 0,
      "votes": 24
    },
    {
      "id": 40059,
      "title": "[LB 0.9971] Full solution with pretrained weights*",
      "comment_count": 9,
      "views": 0,
      "votes": 26
    },
    {
      "id": 40152,
      "title": "My solution with code Public 99.71  but Private 99.67  ",
      "comment_count": 3,
      "views": 0,
      "votes": 10
    },
    {
      "id": 40156,
      "title": "The difference between 1st and 2nd places is astonishing",
      "comment_count": 2,
      "views": 0,
      "votes": 12
    },
    {
      "id": 39897,
      "title": "Data visualization",
      "comment_count": 16,
      "views": 0,
      "votes": 13
    },
    {
      "id": 40092,
      "title": "GPUs are hot at the eleventh hour!",
      "comment_count": 12,
      "views": 0,
      "votes": 5
    },
    {
      "id": 40104,
      "title": "So much silent drama in the last minutes!",
      "comment_count": 7,
      "views": 0,
      "votes": 9
    },
    {
      "id": 40124,
      "title": "Did it overfit?",
      "comment_count": 2,
      "views": 0,
      "votes": 5
    },
    {
      "id": 40172,
      "title": "Summary table of the best solutions",
      "comment_count": 0,
      "views": 0,
      "votes": 7
    },
    {
      "id": 39948,
      "title": "smaller car may gives higher error?",
      "comment_count": 4,
      "views": 0,
      "votes": 9
    },
    {
      "id": 40120,
      "title": "Can't select final submission?",
      "comment_count": 4,
      "views": 0,
      "votes": null
    },
    {
      "id": 39973,
      "title": "Train Dice Coefficient vs Validation DC vs LB",
      "comment_count": 15,
      "views": 0,
      "votes": 1
    },
    {
      "id": 39693,
      "title": "Synthetic data has arrived!",
      "comment_count": 6,
      "views": 0,
      "votes": 43
    },
    {
      "id": 39974,
      "title": "LB shake-up possibility!",
      "comment_count": 3,
      "views": 0,
      "votes": 2
    },
    {
      "id": 39877,
      "title": "After competition reaserch",
      "comment_count": 2,
      "views": 0,
      "votes": 5
    },
    {
      "id": 39566,
      "title": "Dangers of Deep Learning",
      "comment_count": 8,
      "views": 0,
      "votes": 18
    },
    {
      "id": 39690,
      "title": "Hacking GTA V for Carvana Kaggle Challenge – Hacker Noon",
      "comment_count": 6,
      "views": 0,
      "votes": 6
    },
    {
      "id": 39902,
      "title": "Local Validation Loss",
      "comment_count": 1,
      "views": 0,
      "votes": null
    },
    {
      "id": 39383,
      "title": "How to ensemble my several models?",
      "comment_count": 13,
      "views": 0,
      "votes": 3
    },
    {
      "id": 39661,
      "title": "if we ensemble the results of all kagglers with LB>0.9969, what will happen?",
      "comment_count": 9,
      "views": 0,
      "votes": 11
    },
    {
      "id": 39157,
      "title": "some experiments and LB scores",
      "comment_count": 20,
      "views": 0,
      "votes": 35
    },
    {
      "id": 39708,
      "title": "Predicted Masks - No Training",
      "comment_count": 2,
      "views": 0,
      "votes": 7
    },
    {
      "id": 39571,
      "title": "ERROR: File appears to be empty (Line 1, Column 1)",
      "comment_count": 4,
      "views": 0,
      "votes": null
    },
    {
      "id": 39539,
      "title": "RLE decoding?",
      "comment_count": 2,
      "views": 0,
      "votes": 1
    },
    {
      "id": 39285,
      "title": "Review about Semantic Segmentation",
      "comment_count": 6,
      "views": 0,
      "votes": 9
    },
    {
      "id": 39507,
      "title": "How do you resize mask from net output back to original size?",
      "comment_count": 3,
      "views": 0,
      "votes": null
    },
    {
      "id": 39393,
      "title": "image pre-processing",
      "comment_count": 2,
      "views": 0,
      "votes": 6
    },
    {
      "id": 38807,
      "title": "High quality jpg images are now available",
      "comment_count": 12,
      "views": 0,
      "votes": 24
    },
    {
      "id": 39278,
      "title": "Will big input size such as 1280 * 1280 overfit?",
      "comment_count": 7,
      "views": 0,
      "votes": 1
    },
    {
      "id": 39017,
      "title": "Teaming up",
      "comment_count": 15,
      "views": 0,
      "votes": 1
    },
    {
      "id": 39206,
      "title": "Different Activation",
      "comment_count": 13,
      "views": 0,
      "votes": 2
    },
    {
      "id": 39406,
      "title": "Damaged test image in the dataset",
      "comment_count": 2,
      "views": 0,
      "votes": null
    },
    {
      "id": 39348,
      "title": "What is you uNet training time?",
      "comment_count": 9,
      "views": 0,
      "votes": null
    },
    {
      "id": 39358,
      "title": "Did anyone tried doing this in R",
      "comment_count": 2,
      "views": 0,
      "votes": null
    },
    {
      "id": 39148,
      "title": "More batch size or bigger input size?",
      "comment_count": 6,
      "views": 0,
      "votes": null
    },
    {
      "id": 38818,
      "title": "Leaderboard is fixed",
      "comment_count": 18,
      "views": 0,
      "votes": 26
    },
    {
      "id": 39223,
      "title": "Another dead end for Kaggle!",
      "comment_count": 6,
      "views": 0,
      "votes": -5
    },
    {
      "id": 38910,
      "title": "Why it's impossible to use batch=1 with BatchNorm?",
      "comment_count": 3,
      "views": 0,
      "votes": null
    },
    {
      "id": 39060,
      "title": "Running out of time?",
      "comment_count": 7,
      "views": 0,
      "votes": 8
    },
    {
      "id": 38811,
      "title": "any advantage  to use image beyond 1024x1024?",
      "comment_count": 23,
      "views": 0,
      "votes": 6
    },
    {
      "id": 39098,
      "title": "Is this an anomaly in ranking??",
      "comment_count": 7,
      "views": 0,
      "votes": null
    },
    {
      "id": 38879,
      "title": "What augmentation methods do you use?",
      "comment_count": 4,
      "views": 0,
      "votes": 8
    },
    {
      "id": 38633,
      "title": "Visualizing hidden layer activations",
      "comment_count": 7,
      "views": 0,
      "votes": 23
    },
    {
      "id": 38952,
      "title": "Quick Google Scholar Search -> Possible Experiments",
      "comment_count": 0,
      "views": 0,
      "votes": 2
    },
    {
      "id": 38870,
      "title": "How many parameters does your best model have?",
      "comment_count": 1,
      "views": 0,
      "votes": 1
    },
    {
      "id": 38926,
      "title": "predict labels in groups of nxn pixels instead of single pixel",
      "comment_count": 1,
      "views": 0,
      "votes": 5
    },
    {
      "id": 37133,
      "title": "Is deep learning a requirement for this competition?",
      "comment_count": 7,
      "views": 0,
      "votes": 3
    },
    {
      "id": 38359,
      "title": "To Admin,We need a valid and useful Public LB.",
      "comment_count": 35,
      "views": 0,
      "votes": 45
    },
    {
      "id": 37738,
      "title": "Small minibatches? Consider batch renormalization.",
      "comment_count": 10,
      "views": 0,
      "votes": 7
    },
    {
      "id": 38125,
      "title": "how to move to 0.997",
      "comment_count": 61,
      "views": 0,
      "votes": 81
    },
    {
      "id": 38769,
      "title": "should we spam the discussion forum to fix the LB ?!!! ",
      "comment_count": 16,
      "views": 0,
      "votes": -1
    },
    {
      "id": 37243,
      "title": "If your submission file is too large,use 7z format",
      "comment_count": 14,
      "views": 0,
      "votes": 4
    },
    {
      "id": 38812,
      "title": "Submission evaluation",
      "comment_count": 7,
      "views": 0,
      "votes": 2
    },
    {
      "id": 37603,
      "title": "How long does it take to evaluate a single submission?",
      "comment_count": 4,
      "views": 0,
      "votes": null
    },
    {
      "id": 37229,
      "title": "Mask errors and winning strategies",
      "comment_count": 43,
      "views": 0,
      "votes": 19
    },
    {
      "id": 38832,
      "title": "Anyone Else Having Issues With Unicode Encoding in the Image Files?",
      "comment_count": 0,
      "views": 0,
      "votes": 2
    },
    {
      "id": 37681,
      "title": "Trained Unet @ 1918x1280, got LB 99.6 ~ 1 hour per epoch. Drawbacks?",
      "comment_count": 49,
      "views": 0,
      "votes": 22
    },
    {
      "id": 38816,
      "title": "Improving CNN Performance with Lower Resolution Images",
      "comment_count": 3,
      "views": 0,
      "votes": null
    },
    {
      "id": 38703,
      "title": "Reached the edge of my limitations",
      "comment_count": 20,
      "views": 0,
      "votes": 1
    },
    {
      "id": 38478,
      "title": "Imagemagick says that your image quality is 55",
      "comment_count": 18,
      "views": 0,
      "votes": 18
    },
    {
      "id": 38912,
      "title": "Diabetic Retinopathy",
      "comment_count": 1,
      "views": 0,
      "votes": -13
    },
    {
      "id": 38189,
      "title": "from 0.997 to 0.998",
      "comment_count": 19,
      "views": 0,
      "votes": 28
    },
    {
      "id": 38643,
      "title": "How to transfer improvements from 128x128 to 1024x1024?",
      "comment_count": 6,
      "views": 0,
      "votes": 1
    },
    {
      "id": 38503,
      "title": "Uploading is slow",
      "comment_count": 9,
      "views": 0,
      "votes": 1
    },
    {
      "id": 38375,
      "title": "Any advise for CPU users not GPU for this competition?",
      "comment_count": 12,
      "views": 0,
      "votes": null
    },
    {
      "id": 38629,
      "title": "how to speed up image processing for training/testing",
      "comment_count": 0,
      "views": 0,
      "votes": 3
    },
    {
      "id": 38571,
      "title": "What software to use to edit the masks in training sets to correct error",
      "comment_count": 8,
      "views": 0,
      "votes": 1
    },
    {
      "id": 37240,
      "title": "Is it legal to make use of TEST SET for unsupervised learning?",
      "comment_count": 3,
      "views": 0,
      "votes": 4
    },
    {
      "id": 38301,
      "title": "Exact public score in HTML source code",
      "comment_count": 14,
      "views": 0,
      "votes": 35
    },
    {
      "id": 38547,
      "title": "another benchmark with tensorboard visualization",
      "comment_count": 0,
      "views": 0,
      "votes": 8
    },
    {
      "id": 38534,
      "title": "My worst predictions.",
      "comment_count": 6,
      "views": 0,
      "votes": 7
    },
    {
      "id": 38298,
      "title": "initial results on pseudo labeling",
      "comment_count": 9,
      "views": 0,
      "votes": 14
    },
    {
      "id": 38413,
      "title": "Pre-trained models",
      "comment_count": 1,
      "views": 0,
      "votes": null
    },
    {
      "id": 37548,
      "title": "My submission file is too large",
      "comment_count": 3,
      "views": 0,
      "votes": null
    },
    {
      "id": 38261,
      "title": "Using Metadata with CNNs",
      "comment_count": 9,
      "views": 0,
      "votes": 7
    },
    {
      "id": 38354,
      "title": "Evaluation Exception: Stream was not readable.",
      "comment_count": 2,
      "views": 0,
      "votes": null
    },
    {
      "id": 38349,
      "title": "What did not go well with your model?",
      "comment_count": 0,
      "views": 0,
      "votes": 0
    },
    {
      "id": 38238,
      "title": "testing on non kaggle images",
      "comment_count": 2,
      "views": 0,
      "votes": 2
    },
    {
      "id": 38240,
      "title": "is it possible to complete this challenge , without downloading any file on local machine. ",
      "comment_count": 3,
      "views": 0,
      "votes": null
    },
    {
      "id": 38051,
      "title": "Corrupted train masks",
      "comment_count": 6,
      "views": 0,
      "votes": 15
    },
    {
      "id": 38246,
      "title": "LB Scores",
      "comment_count": 2,
      "views": 0,
      "votes": null
    },
    {
      "id": 38226,
      "title": "Resolution",
      "comment_count": 1,
      "views": 0,
      "votes": -2
    },
    {
      "id": 38034,
      "title": "Model heartshock: Is this the so-called catastrophic forgetting?",
      "comment_count": 4,
      "views": 0,
      "votes": -1
    },
    {
      "id": 37968,
      "title": "Crawling Carvana website for more training images ... is it against the rule?",
      "comment_count": 6,
      "views": 0,
      "votes": null
    },
    {
      "id": 38118,
      "title": "Has anybody tried different networks structures in downsampling?",
      "comment_count": 0,
      "views": 0,
      "votes": 0
    },
    {
      "id": 38007,
      "title": "Car color distribution. Blue car advantage?",
      "comment_count": 1,
      "views": 0,
      "votes": 5
    },
    {
      "id": 37808,
      "title": "Weighted DICE loss",
      "comment_count": 6,
      "views": 0,
      "votes": 4
    },
    {
      "id": 38017,
      "title": "Pre-trained model weights or CRF post-processing or atrous convolutions/deeplabv3 or PSPNet?",
      "comment_count": 2,
      "views": 0,
      "votes": 1
    },
    {
      "id": 37958,
      "title": "Training Image rescaling",
      "comment_count": 2,
      "views": 0,
      "votes": null
    },
    {
      "id": 37966,
      "title": "Does using a cloud computing service is allowed?",
      "comment_count": 3,
      "views": 0,
      "votes": null
    },
    {
      "id": 37819,
      "title": "David Güera first submission ever, 3rd position",
      "comment_count": 8,
      "views": 0,
      "votes": -5
    },
    {
      "id": 37800,
      "title": "The real test set has the same size as the train set!",
      "comment_count": 0,
      "views": 0,
      "votes": 7
    },
    {
      "id": 37890,
      "title": "The pixels are numbered from top to bottom, then left to right?",
      "comment_count": 2,
      "views": 0,
      "votes": null
    },
    {
      "id": 37737,
      "title": "Unsupervised pre-training, does it make sense here?",
      "comment_count": 2,
      "views": 0,
      "votes": 2
    },
    {
      "id": 37578,
      "title": "survey on input image size vs LB score",
      "comment_count": 4,
      "views": 0,
      "votes": 10
    },
    {
      "id": 37307,
      "title": "One thought - Using GAN to generate the mask",
      "comment_count": 15,
      "views": 0,
      "votes": 3
    },
    {
      "id": 37659,
      "title": "Glasses",
      "comment_count": 1,
      "views": 0,
      "votes": 1
    },
    {
      "id": 37247,
      "title": "Corrupted image in test set",
      "comment_count": 7,
      "views": 0,
      "votes": 6
    },
    {
      "id": 37611,
      "title": "best way to upscale images",
      "comment_count": 0,
      "views": 0,
      "votes": 2
    },
    {
      "id": 37096,
      "title": "Any advice/tips on how to solve this problem?",
      "comment_count": 9,
      "views": 0,
      "votes": 8
    },
    {
      "id": 37467,
      "title": "Submission file format in overview section is incorrect.",
      "comment_count": 4,
      "views": 0,
      "votes": 7
    },
    {
      "id": 37584,
      "title": "Submission error: Overlapping values detected. Submitted values must be unique.",
      "comment_count": 0,
      "views": 0,
      "votes": 0
    },
    {
      "id": 37455,
      "title": "Photo without a car",
      "comment_count": 5,
      "views": 0,
      "votes": 3
    },
    {
      "id": 37496,
      "title": "How to save RLE string to a gif file to get the mask？",
      "comment_count": 1,
      "views": 0,
      "votes": null
    },
    {
      "id": 37264,
      "title": "Using computer graphics to render car training samples",
      "comment_count": 5,
      "views": 0,
      "votes": 5
    },
    {
      "id": 37314,
      "title": "RLE encoding axes order",
      "comment_count": 3,
      "views": 0,
      "votes": 8
    },
    {
      "id": 37048,
      "title": "Know your car to build better models!",
      "comment_count": 18,
      "views": 0,
      "votes": 11
    },
    {
      "id": 37313,
      "title": "Interesting fact about Carvana",
      "comment_count": 0,
      "views": 0,
      "votes": 4
    },
    {
      "id": 37377,
      "title": "How long does BN take to converge?",
      "comment_count": 1,
      "views": 0,
      "votes": 1
    },
    {
      "id": 37075,
      "title": "Carvana's security rules",
      "comment_count": 17,
      "views": 0,
      "votes": 7
    },
    {
      "id": 37273,
      "title": "alternative methods not using deep learning CNN",
      "comment_count": 1,
      "views": 0,
      "votes": 12
    },
    {
      "id": 37131,
      "title": "Leaderboard Evaluation Dice decimals",
      "comment_count": 3,
      "views": 0,
      "votes": 5
    },
    {
      "id": 37203,
      "title": "how to speed up csv generation?",
      "comment_count": 1,
      "views": 0,
      "votes": 3
    },
    {
      "id": 37281,
      "title": "Can we seperately work on each angle?",
      "comment_count": 1,
      "views": 0,
      "votes": null
    },
    {
      "id": 37092,
      "title": "Similar car models in train and test datasets",
      "comment_count": 5,
      "views": 0,
      "votes": 9
    },
    {
      "id": 37116,
      "title": "Submission file size",
      "comment_count": 4,
      "views": 0,
      "votes": 2
    },
    {
      "id": 37125,
      "title": "Why Carnava does not use green screen technique in their studios?",
      "comment_count": 1,
      "views": 0,
      "votes": 9
    },
    {
      "id": 37169,
      "title": "mask file dimensions  and gif format",
      "comment_count": 4,
      "views": 0,
      "votes": 1
    },
    {
      "id": 37183,
      "title": "Test zip file size",
      "comment_count": 1,
      "views": 0,
      "votes": 1
    },
    {
      "id": 37178,
      "title": "Background subtraction libraries and algorithms",
      "comment_count": 1,
      "views": 0,
      "votes": 1
    },
    {
      "id": 37108,
      "title": "Are Image Coordinates Zero-Indexed or One-Indexed?",
      "comment_count": 2,
      "views": 0,
      "votes": 3
    },
    {
      "id": 37163,
      "title": "Loss function unstable",
      "comment_count": 2,
      "views": 0,
      "votes": null
    }
  ],
  "errors": []
}