![\"0\"](\"https://i.ibb.co/C7qybnC/4500c3929417e8.jpg\")
This topic is just for fun (my goal was to show some anomalies in dataset - as a result we have to address this during solution pipeline definition).
\nMost of EDAs published here show … \"difficult\" photos but they are rather with difficult background - whale is recognizable in my opinion. Below you can find whales which are impossible to recognize. What do you think?
\nDo you recognize me? :)
\n
![\"1\"](\"https://i.ibb.co/wzCSbNG/00eb3441343226.jpg\")
![\"2\"](\"https://i.ibb.co/y6qVV5m/1a58dce81de5e6.jpg\")
![\"3\"](\"https://i.ibb.co/Z1GBTmD/1a761a0620e2ad.jpg\")
![\"4\"](\"https://i.ibb.co/gT08jqd/1ad22afb248f51.jpg\")
![\"5\"](\"https://i.ibb.co/CKhDzB5/1cc8ddb84be688.jpg\")
![\"6\"](\"https://i.ibb.co/47XB219/1f1aae217eb5ee.jpg\")
![\"7\"](\"https://i.ibb.co/Kwq1ywT/02ab9cd7831474.jpg\")
![\"8\"](\"https://i.ibb.co/WBdgS1d/26962189917f2e.jpg\")
![\"9\"](\"https://i.ibb.co/fpRSTN6/2e966e115143e1.jpg\")
![\"10\"](\"https://i.ibb.co/nLGCVhw/2f9c77a0e4ffa8.jpg\")
![\"11\"](\"https://i.ibb.co/ThqYxC6/12fd305f641c4d.jpg\")
![\"12\"](\"https://i.ibb.co/QdXB67c/15d5ee1ffcfec7.jpg\")
![\"13\"](\"https://i.ibb.co/vVxvYTK/21fb805992d396.jpg\")
![\"14\"](\"https://i.ibb.co/BG99hpb/056c1b6d4fe5f1.jpg\")
![\"15\"](\"https://i.ibb.co/GMWZwWy/13484190d4f35d.jpg\")
![\"16\"](\"https://i.ibb.co/XSVWdCQ/227afed8c4d971.jpg\")
![\"17\"](\"https://i.ibb.co/rdXqhH8/234c3b7e517e86.jpg\")
![\"18\"](\"https://i.ibb.co/zZg9yzd/2067db1f3ecbdb.jpg\")
![\"19\"](\"https://i.ibb.co/TPY3ngS/20055b3c3a7214.jpg\")
![\"20\"](\"https://i.ibb.co/SXJ3xJn/084318fd704995.jpg\")
![\"21\"](\"https://i.ibb.co/hWDMPnY/09d88e169e2b40.jpg\")
This topic is just for fun (my goal was to show some anomalies in dataset - as a result we have to address this during solution pipeline definition).
\nMost of EDAs published here show … \"difficult\" photos but they are rather with difficult background - whale is recognizable in my opinion. Below you can find whales which are impossible to recognize. What do you think?
\nDo you recognize me? :)
\n
I wonder how they annotate this dataset, they are using electronic chips?
", "rawMarkdown": "I wonder how they annotate this dataset, they are using electronic chips?", "votes": null }, { "id": "1700666", "postDate": "02/22/2022 06:20:19", "content": "One way is to chip but another way is to get closer and make next photo - this post is a tip to check it in dataset … do not to recognize whale (in test) based on rubbish input data.
", "rawMarkdown": "One way is to chip but another way is to get closer and make next photo - this post is a tip to check it in dataset ... do not to recognize whale (in test) based on rubbish input data.", "votes": null }, { "id": "1700674", "postDate": "02/22/2022 06:26:57", "content": "The competition makes me feel like making an ID pool. The pictures in the ID pool represent different parts of the same ID. Then the more diverse the ID pool is, the greater the scale difference is, and it can be identified from different angles
", "rawMarkdown": "The competition makes me feel like making an ID pool. The pictures in the ID pool represent different parts of the same ID. Then the more diverse the ID pool is, the greater the scale difference is, and it can be identified from different angles", "votes": null }, { "id": "1700695", "postDate": "02/22/2022 07:03:53", "content": "Exactly! Unfortunately we have many single image cases 😨 But we will deal with this as well (maybe …) 🤕🤒😲🤥😆
", "rawMarkdown": "Exactly! Unfortunately we have many single image cases 😨 But we will deal with this as well (maybe ...) 🤕🤒😲🤥😆", "votes": null }, { "id": "1700727", "postDate": "02/22/2022 07:38:14", "content": "![\"image\"](\"https://s2.loli.net/2022/02/22/vykFYRWc3aZoVq5.png\")
\nX is the stage of training image number, Y is validation score.
\nIt is very important to find out why some single picture score is better
The same whale ID?
", "rawMarkdown": "The same whale ID?", "votes": null }, { "id": "1700806", "postDate": "02/22/2022 09:14:07", "content": "Have you done dataset cleaning? Look here: https://www.kaggle.com/c/happy-whale-and-dolphin/discussion/309188
", "rawMarkdown": "Have you done dataset cleaning? Look here: https://www.kaggle.com/c/happy-whale-and-dolphin/discussion/309188", "votes": null }, { "id": "1704530", "postDate": "02/25/2022 15:50:44", "content": "No, I am focusing on how to make the network extract the distinguishing local features in the image
", "rawMarkdown": "No, I am focusing on how to make the network extract the distinguishing local features in the image", "votes": null }, { "id": "1704713", "postDate": "02/25/2022 18:19:23", "content": "I'm no expert, but my girlfriend is a marine biologist. From what she told me, when spotting they usually have \"big eyes\" binoculars on the ship. So my guess is they can get enough detail to identify with big eyes, but the camera used to take the picture can not. Just my guess.
\nAs far as annotation, in this I would assume it's manual.
Thank you for providing this information. On Kaggle I learn new thnings everyday - even they are not directly conneted with ML/AI :)
", "rawMarkdown": "Thank you for providing this information. On Kaggle I learn new thnings everyday - even they are not directly conneted with ML/AI :)", "votes": null } ], "comments": [ { "id": 1700468, "author_name": "ptran1203", "author_url": "", "post_date": "02/22/2022 02:56:01", "content": "I wonder how they annotate this dataset, they are using electronic chips?
", "votes": null, "replies": [ { "id": 1700666, "author_name": "remekkinas", "author_url": "", "post_date": "02/22/2022 06:20:19", "content": "One way is to chip but another way is to get closer and make next photo - this post is a tip to check it in dataset … do not to recognize whale (in test) based on rubbish input data.
", "votes": null, "replies": [] }, { "id": 1704713, "author_name": "click2clone", "author_url": "", "post_date": "02/25/2022 18:19:23", "content": "I'm no expert, but my girlfriend is a marine biologist. From what she told me, when spotting they usually have \"big eyes\" binoculars on the ship. So my guess is they can get enough detail to identify with big eyes, but the camera used to take the picture can not. Just my guess.
\nAs far as annotation, in this I would assume it's manual.
Thank you for providing this information. On Kaggle I learn new thnings everyday - even they are not directly conneted with ML/AI :)
", "votes": null, "replies": [] } ] }, { "id": 1700674, "author_name": "biglafe", "author_url": "", "post_date": "02/22/2022 06:26:57", "content": "The competition makes me feel like making an ID pool. The pictures in the ID pool represent different parts of the same ID. Then the more diverse the ID pool is, the greater the scale difference is, and it can be identified from different angles
", "votes": null, "replies": [ { "id": 1700695, "author_name": "remekkinas", "author_url": "", "post_date": "02/22/2022 07:03:53", "content": "Exactly! Unfortunately we have many single image cases 😨 But we will deal with this as well (maybe …) 🤕🤒😲🤥😆
", "votes": null, "replies": [] }, { "id": 1700727, "author_name": "biglafe", "author_url": "", "post_date": "02/22/2022 07:38:14", "content": "
\nX is the stage of training image number, Y is validation score.
\nIt is very important to find out why some single picture score is better
The same whale ID?
", "votes": null, "replies": [] }, { "id": 1700806, "author_name": "remekkinas", "author_url": "", "post_date": "02/22/2022 09:14:07", "content": "Have you done dataset cleaning? Look here: https://www.kaggle.com/c/happy-whale-and-dolphin/discussion/309188
", "votes": null, "replies": [] }, { "id": 1704530, "author_name": "biglafe", "author_url": "", "post_date": "02/25/2022 15:50:44", "content": "No, I am focusing on how to make the network extract the distinguishing local features in the image
", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "1700350": "This topic is just for fun (my goal was to show some anomalies in dataset - as a result we have to address this during solution pipeline definition). \n\nMost of EDAs published here show ... \"difficult\" photos but they are rather with difficult background - whale is recognizable in my opinion. Below you can find whales which are impossible to recognize. What do you think?\n\n\nDo you recognize me? :) \n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n", "1700468": "I wonder how they annotate this dataset, they are using electronic chips?", "1700666": "One way is to chip but another way is to get closer and make next photo - this post is a tip to check it in dataset ... do not to recognize whale (in test) based on rubbish input data.", "1700674": "The competition makes me feel like making an ID pool. The pictures in the ID pool represent different parts of the same ID. Then the more diverse the ID pool is, the greater the scale difference is, and it can be identified from different angles", "1700695": "Exactly! Unfortunately we have many single image cases 😨 But we will deal with this as well (maybe ...) 🤕🤒😲🤥😆", "1700727": " \nX is the stage of training image number, Y is validation score.\nIt is very important to find out why some single picture score is better", "1700730": "The same whale ID?", "1700806": "Have you done dataset cleaning? Look here: https://www.kaggle.com/c/happy-whale-and-dolphin/discussion/309188", "1704530": "No, I am focusing on how to make the network extract the distinguishing local features in the image", "1704713": "I'm no expert, but my girlfriend is a marine biologist. From what she told me, when spotting they usually have \"big eyes\" binoculars on the ship. So my guess is they can get enough detail to identify with big eyes, but the camera used to take the picture can not. Just my guess.\nAs far as annotation, in this I would assume it's manual.", "1704846": "Thank you for providing this information. On Kaggle I learn new thnings everyday - even they are not directly conneted with ML/AI :)" }, "source": "meta" }