{ "id": 172472, "title": "Thank you to all and rank \"997th\"'s approach", "url": "/competitions/global-wheat-detection/discussion/172472", "author_name": "DeepUnderstanding", "post_date": "2020-08-05T07:09:55.184000", "votes": 6, "comment_count": 1, "views": 0, "content": "

I want to say thank you to all the people giving awesome notebooks with their very cool Insights.this was my first Kaggle competition so it is special for me. Before joining this competition I knew Object detection only by theory and I was \"Keras Guy\", this competition helped me to implement a full model written in Pytorch and I am very much happy. Those starter notebooks and discussions were really beneficial.
\nI wanted to share my approach, although it's not that good( you can see the rank and tell). I faced some problems as I have a Laptop with only 4gb of RAM and 2gb GPU, I had to solely depend on Kaggle kernels, which had a limited time, doing it was a little hard for me.
\nSo here is my approach:

\n
    \n
  1. I used 512X512 sized images to train a base FasterRCNN(resnet50) model using all the data and no data augmentation (30 epochs).
    2. \n
  2. I used Stratified KFolds to make 5 folds of the data.
    3. \n
  3. Each fold of data had the same augmentations including horizontal vertical flipping, Random Shadow, Gaussian Blur, Cutouts.
    4. \n
  4. I used the base model to retrain on each fold now using 1024x1024 images with augmentations.
    5. \n
  5. Lastly, I used WBF ensemble to get the result from each of the models.
    \nI will be really happy if you comment hoe could I make this model better or why didn't it worked.
    \nSo this was In my part, I hope I can use the knowledge in other competitions, hope you guys are safe and safe and see you soon in other competitions. I am so eager to see top solutions!!
    \nSpecial thanks to @pestipeti @kaushal2896 @matthewmasters for the notebooks and discussions and clearing my doubts.
    \nThankyou!!!😀
    6. \n
", "messages": [ { "id": 958845, "postDate": "2020-08-05T07:09:55.183Z", "content": "

I want to say thank you to all the people giving awesome notebooks with their very cool Insights.this was my first Kaggle competition so it is special for me. Before joining this competition I knew Object detection only by theory and I was \"Keras Guy\", this competition helped me to implement a full model written in Pytorch and I am very much happy. Those starter notebooks and discussions were really beneficial.
\nI wanted to share my approach, although it's not that good( you can see the rank and tell). I faced some problems as I have a Laptop with only 4gb of RAM and 2gb GPU, I had to solely depend on Kaggle kernels, which had a limited time, doing it was a little hard for me.
\nSo here is my approach:

\n
    \n
  1. I used 512X512 sized images to train a base FasterRCNN(resnet50) model using all the data and no data augmentation (30 epochs).
    2. \n
  2. I used Stratified KFolds to make 5 folds of the data.
    3. \n
  3. Each fold of data had the same augmentations including horizontal vertical flipping, Random Shadow, Gaussian Blur, Cutouts.
    4. \n
  4. I used the base model to retrain on each fold now using 1024x1024 images with augmentations.
    5. \n
  5. Lastly, I used WBF ensemble to get the result from each of the models.
    \nI will be really happy if you comment hoe could I make this model better or why didn't it worked.
    \nSo this was In my part, I hope I can use the knowledge in other competitions, hope you guys are safe and safe and see you soon in other competitions. I am so eager to see top solutions!!
    \nSpecial thanks to @pestipeti @kaushal2896 @matthewmasters for the notebooks and discussions and clearing my doubts.
    \nThankyou!!!😀
    6. \n
", "rawMarkdown": "I want to say thank you to all the people giving awesome notebooks with their very cool Insights.this was my first Kaggle competition so it is special for me. Before joining this competition I knew Object detection only by theory and I was \"Keras Guy\", this competition helped me to implement a full model written in Pytorch and I am very much happy. Those starter notebooks and discussions were really beneficial.\nI wanted to share my approach, although it's not that good( you can see the rank and tell). I faced some problems as I have a Laptop with only 4gb of RAM and 2gb GPU, I had to solely depend on Kaggle kernels, which had a limited time, doing it was a little hard for me.\nSo here is my approach:\n1. I used 512X512 sized images to train a base FasterRCNN(resnet50) model using all the data and no data augmentation (30 epochs).\n2. I used Stratified KFolds to make 5 folds of the data.\n3. Each fold of data had the same augmentations including horizontal vertical flipping, Random Shadow, Gaussian Blur, Cutouts.\n4. I used the base model to retrain on each fold now using 1024x1024 images with augmentations.\n5. Lastly, I used WBF ensemble to get the result from each of the models.\nI will be really happy if you comment hoe could I make this model better or why didn't it worked.\nSo this was In my part, I hope I can use the knowledge in other competitions, hope you guys are safe and safe and see you soon in other competitions. I am so eager to see top solutions!!\nSpecial thanks to @pestipeti @kaushal2896 @matthewmasters for the notebooks and discussions and clearing my doubts.\nThankyou!!!😀 ", "votes": 6 }, { "id": 958937, "postDate": "2020-08-05T08:22:03.827Z", "rawMarkdown": "", "isDeleted": true } ], "comments": [ { "id": 958937, "author_name": "", "author_url": "", "post_date": "2020-08-05T08:22:03.827000", "content": "", "votes": 0, "replies": [] } ], "raw_markdown_by_id": { "958845": "I want to say thank you to all the people giving awesome notebooks with their very cool Insights.this was my first Kaggle competition so it is special for me. Before joining this competition I knew Object detection only by theory and I was \"Keras Guy\", this competition helped me to implement a full model written in Pytorch and I am very much happy. Those starter notebooks and discussions were really beneficial.\nI wanted to share my approach, although it's not that good( you can see the rank and tell). I faced some problems as I have a Laptop with only 4gb of RAM and 2gb GPU, I had to solely depend on Kaggle kernels, which had a limited time, doing it was a little hard for me.\nSo here is my approach:\n1. I used 512X512 sized images to train a base FasterRCNN(resnet50) model using all the data and no data augmentation (30 epochs).\n2. I used Stratified KFolds to make 5 folds of the data.\n3. Each fold of data had the same augmentations including horizontal vertical flipping, Random Shadow, Gaussian Blur, Cutouts.\n4. I used the base model to retrain on each fold now using 1024x1024 images with augmentations.\n5. Lastly, I used WBF ensemble to get the result from each of the models.\nI will be really happy if you comment hoe could I make this model better or why didn't it worked.\nSo this was In my part, I hope I can use the knowledge in other competitions, hope you guys are safe and safe and see you soon in other competitions. I am so eager to see top solutions!!\nSpecial thanks to @pestipeti @kaushal2896 @matthewmasters for the notebooks and discussions and clearing my doubts.\nThankyou!!!😀 ", "958937": "" } }