I am noob. I don't have any prior knowledge of images.
\nI taken one sample from our data set and tried to view that using matplotlib's plt.imshow() and i used only 1 frame, because that .npy was shape of hxwxT and T which is 8 and we can't use 8 we can maximum use 3 for rgb. And my image is looking different![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F6046134%2Fafc189daa20adfda72af91c40fb01eb8%2FScreenshot%20(10).png?generation=1684491377989029&alt=media\")
\nand i want something like this (colorful)
\n![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F6046134%2F3ff55bc01e8b12093b661ae26339be32%2Finbox_59561_590a0bc76044a4ceb71368cf3b62412e_contrails_600.png?generation=1684491595022625&alt=media\")
Hi Arvind,
\nThere is example code for how to produce the false-color images at https://www.kaggle.com/code/inversion/visualizing-contrails?scriptVersionId=129621747&cellId=4
", "rawMarkdown": "Hi Arvind,\n\nThere is example code for how to produce the false-color images at https://www.kaggle.com/code/inversion/visualizing-contrails?scriptVersionId=129621747&cellId=4", "votes": 1, "replies": [ { "id": 2265831, "postDate": "2023-05-19T14:42:17.747Z", "content": "oh. Thank you.
", "rawMarkdown": "oh. Thank you." }, { "id": 2266642, "postDate": "2023-05-20T08:36:55.770Z", "content": "Hi, i saw we used band 11, band 14, and band 15 to form image but we didn't used other combination like 8, 9, 10 or 8, 10, 12 and the normalization technique we used in this that was not normal
", "rawMarkdown": "Hi, i saw we used band 11, band 14, and band 15 to form image but we didn't used other combination like 8, 9, 10 or 8, 10, 12 and the normalization technique we used in this that was not normal", "replies": [ { "id": 2267471, "postDate": "2023-05-21T01:21:06.727Z", "content": "This is purely a visualization technique tuned to human perception, and it's likely not optimal even for that. There could be other useful information in the other bands that help a neural network identify contrails better than humans can. It's also very standard to use different data normalization for image visualization vs input to a neural network. All that said, this was the visualization used by the humans who labeled the dataset.
", "rawMarkdown": "This is purely a visualization technique tuned to human perception, and it's likely not optimal even for that. There could be other useful information in the other bands that help a neural network identify contrails better than humans can. It's also very standard to use different data normalization for image visualization vs input to a neural network. All that said, this was the visualization used by the humans who labeled the dataset.", "votes": 2 } ] } ] }, { "id": 2265541, "postDate": "2023-05-19T10:22:11.353Z", "content": "I am noob. I don't have any prior knowledge of images.
\nI taken one sample from our data set and tried to view that using matplotlib's plt.imshow() and i used only 1 frame, because that .npy was shape of hxwxT and T which is 8 and we can't use 8 we can maximum use 3 for rgb. And my image is looking different
\nand i want something like this (colorful)
\n
You can check RGB or GBR first when using imshow(), hope it can help you.
", "rawMarkdown": "You can check RGB or GBR first when using imshow(), hope it can help you.", "votes": 1, "replies": [ { "id": 2265550, "postDate": "2023-05-19T10:30:57.017Z", "content": "is it possible to use rgb , because my image dim is HxWx1
", "rawMarkdown": "is it possible to use rgb , because my image dim is HxWx1", "votes": 1 } ] }, { "id": 2265795, "postDate": "2023-05-19T14:11:06.907Z", "content": "I think T is the number of frames. There are nine bands. RGB here does not make sense other than pretty pictures.
\nTry R = band_08, G = band_12 and B = band_16 ?
Hi Arvind,
\nThere is example code for how to produce the false-color images at https://www.kaggle.com/code/inversion/visualizing-contrails?scriptVersionId=129621747&cellId=4
", "votes": 1, "replies": [ { "id": 2265831, "author_name": "Arvind Devarkonda", "author_url": "", "post_date": "2023-05-19T14:42:17.747000", "content": "oh. Thank you.
", "votes": 0, "replies": [] }, { "id": 2266642, "author_name": "Arvind Devarkonda", "author_url": "", "post_date": "2023-05-20T08:36:55.770000", "content": "Hi, i saw we used band 11, band 14, and band 15 to form image but we didn't used other combination like 8, 9, 10 or 8, 10, 12 and the normalization technique we used in this that was not normal
", "votes": 0, "replies": [ { "id": 2267471, "author_name": "Aaron Sarna", "author_url": "", "post_date": "2023-05-21T01:21:06.727000", "content": "This is purely a visualization technique tuned to human perception, and it's likely not optimal even for that. There could be other useful information in the other bands that help a neural network identify contrails better than humans can. It's also very standard to use different data normalization for image visualization vs input to a neural network. All that said, this was the visualization used by the humans who labeled the dataset.
", "votes": 2, "replies": [] } ] } ] }, { "id": 2265542, "author_name": "Dewei Chen", "author_url": "", "post_date": "2023-05-19T10:24:27.170000", "content": "You can check RGB or GBR first when using imshow(), hope it can help you.
", "votes": 1, "replies": [ { "id": 2265550, "author_name": "Arvind Devarkonda", "author_url": "", "post_date": "2023-05-19T10:30:57.017000", "content": "is it possible to use rgb , because my image dim is HxWx1
", "votes": 1, "replies": [] } ] }, { "id": 2265795, "author_name": "Tord Malmgren", "author_url": "", "post_date": "2023-05-19T14:11:06.907000", "content": "I think T is the number of frames. There are nine bands. RGB here does not make sense other than pretty pictures.
\nTry R = band_08, G = band_12 and B = band_16 ?