![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F172860%2Fa991912d0f783a10b9e71b1689dfd577%2FScreenshot%20from%202023-08-25%2008-35-11.png?generation=1692945806175845&alt=media\")
I have joined quite recently and even though I didn't get any medal, I managed to learn a few things:
\nFinally, I got the last missing discussion gold and became a discussion GM. Thanks to all for the upvotes! 🥳
\n
Time to read the top solutions now. 🫡
", "messages": [ { "id": 2407610, "postDate": "2023-08-25T06:47:35.563Z", "content": "I have joined quite recently and even though I didn't get any medal, I managed to learn a few things:
\nFinally, I got the last missing discussion gold and became a discussion GM. Thanks to all for the upvotes! 🥳
\n
Time to read the top solutions now. 🫡
", "rawMarkdown": "I have joined quite recently and even though I didn't get any medal, I managed to learn a few things:\n\n* Different **landmarks** can be extracted using the MediaPipe [API](https://developers.google.com/mediapipe).\n* Last ASL top solution is a great starting point \n* I managed to make an **animated plot** of the extracted landmarks given a Parquet file. I will try to share this code at some point (after cleaning it).\n* I haven't used **Tensorflow** on GPU for a very long time. It was tough making it run locally but I managed after reading various posts. It seems the key ingredient was [**modprobe**](https://linux.die.net/man/8/modprobe). I have also managed to make it run using the Docker GPU [image](https://hub.docker.com/r/tensorflow/tensorflow/). \n* Learned many things about the [**CTC loss**](https://paperswithcode.com/method/ctc-loss). This blog post is a very good [explanation](https://distill.pub/2017/ctc/).\n\n\nFinally, I got the last missing discussion gold and became a discussion GM. Thanks to all for the upvotes! 🥳\n\n\n\n\nTime to read the top solutions now. 🫡", "votes": 11 }, { "id": 2408320, "postDate": "2023-08-25T15:11:56.480Z", "content": "Congratulations to you. I think this competition is a great learning opportunity.
\nIt also required us to factor in the notebook time while adding extra layers in processing.
\nEven though we had 5 submissions per day, hardly could submit 3-4 notebooks per week due to GPU limit.
Congratulations @yassinealouini becoming Kaggle discussion grandmaster! Amazing achievement. Than you for sharing all your discussions!
", "rawMarkdown": "Congratulations @yassinealouini becoming Kaggle discussion grandmaster! Amazing achievement. Than you for sharing all your discussions!", "votes": 1, "replies": [ { "id": 2408381, "postDate": "2023-08-25T15:51:20.403Z", "content": "Thanks @cdeotte, means a lot coming from you.
", "rawMarkdown": "Thanks @cdeotte, means a lot coming from you. " } ] }, { "id": 2408038, "postDate": "2023-08-25T11:56:25.617Z", "content": "Wow! very nice @yassinealouini! congratulations!
", "rawMarkdown": "Wow! very nice @yassinealouini! congratulations!", "votes": 1 }, { "id": 2407863, "postDate": "2023-08-25T09:36:53.293Z", "content": "Congratulations for your GM rank
", "rawMarkdown": "Congratulations for your GM rank", "votes": 1, "replies": [ { "id": 2407935, "postDate": "2023-08-25T10:37:25.777Z", "content": "Thanks and congratulation on the great position. 👍
", "rawMarkdown": "Thanks and congratulation on the great position. 👍", "votes": 1 } ] } ], "comments": [ { "id": 2408320, "author_name": "C R Suthikshn Kumar", "author_url": "", "post_date": "2023-08-25T15:11:56.480000", "content": "Congratulations to you. I think this competition is a great learning opportunity.
\nIt also required us to factor in the notebook time while adding extra layers in processing.
\nEven though we had 5 submissions per day, hardly could submit 3-4 notebooks per week due to GPU limit.
Congratulations @yassinealouini becoming Kaggle discussion grandmaster! Amazing achievement. Than you for sharing all your discussions!
", "votes": 1, "replies": [ { "id": 2408381, "author_name": "Yassine Alouini", "author_url": "", "post_date": "2023-08-25T15:51:20.403000", "content": "Thanks @cdeotte, means a lot coming from you.
", "votes": 0, "replies": [] } ] }, { "id": 2408038, "author_name": "Arvind (Yetirajan) Narayanan Iyengar", "author_url": "", "post_date": "2023-08-25T11:56:25.617000", "content": "Wow! very nice @yassinealouini! congratulations!
", "votes": 1, "replies": [] }, { "id": 2407863, "author_name": "Chakkrit Termritthikun", "author_url": "", "post_date": "2023-08-25T09:36:53.293000", "content": "Congratulations for your GM rank
", "votes": 1, "replies": [ { "id": 2407935, "author_name": "Yassine Alouini", "author_url": "", "post_date": "2023-08-25T10:37:25.777000", "content": "Thanks and congratulation on the great position. 👍
", "votes": 1, "replies": [] } ] } ], "raw_markdown_by_id": { "2407610": "I have joined quite recently and even though I didn't get any medal, I managed to learn a few things:\n\n* Different **landmarks** can be extracted using the MediaPipe [API](https://developers.google.com/mediapipe).\n* Last ASL top solution is a great starting point \n* I managed to make an **animated plot** of the extracted landmarks given a Parquet file. I will try to share this code at some point (after cleaning it).\n* I haven't used **Tensorflow** on GPU for a very long time. It was tough making it run locally but I managed after reading various posts. It seems the key ingredient was [**modprobe**](https://linux.die.net/man/8/modprobe). I have also managed to make it run using the Docker GPU [image](https://hub.docker.com/r/tensorflow/tensorflow/). \n* Learned many things about the [**CTC loss**](https://paperswithcode.com/method/ctc-loss). This blog post is a very good [explanation](https://distill.pub/2017/ctc/).\n\n\nFinally, I got the last missing discussion gold and became a discussion GM. Thanks to all for the upvotes! 🥳\n\n\n\n\nTime to read the top solutions now. 🫡", "2408320": "Congratulations to you. I think this competition is a great learning opportunity.\nIt also required us to factor in the notebook time while adding extra layers in processing.\nEven though we had 5 submissions per day, hardly could submit 3-4 notebooks per week due to GPU limit.", "2408243": "Congratulations @yassinealouini becoming Kaggle discussion grandmaster! Amazing achievement. Than you for sharing all your discussions!", "2408038": "Wow! very nice @yassinealouini! congratulations!", "2407863": "Congratulations for your GM rank" } }