I was going through a couple of notebooks to understand the dataset and to find any hint. I was intrigued by the false color image generated using bands 11 -14. But I keep coming back to bands from 8-16.
\nIs building a model using false color image viable or do we need to use all frames ? Any tips will be really helpful.
", "messages": [ { "id": 2260531, "postDate": "2023-05-15T17:47:29.060Z", "content": "I was going through a couple of notebooks to understand the dataset and to find any hint. I was intrigued by the false color image generated using bands 11 -14. But I keep coming back to bands from 8-16.
\nIs building a model using false color image viable or do we need to use all frames ? Any tips will be really helpful.
", "rawMarkdown": "I was going through a couple of notebooks to understand the dataset and to find any hint. I was intrigued by the false color image generated using bands 11 -14. But I keep coming back to bands from 8-16. \n\nIs building a model using false color image viable or do we need to use all frames ? Any tips will be really helpful.", "votes": 5 }, { "id": 2260585, "postDate": "2023-05-15T18:31:14.347Z", "content": "Hi Kenni,
\nI'm actually asking myself the same question. In doubt, I'm going to go with all the bands (and the network will do its job), but as time is a (big) issue, maybe discarding some bands is the way to go…
\nI'm not very helpful here, just letting you know I'm in the same problematic.
\n\nHi Kenni,
\n
\n I'm actually asking myself the same question. In doubt, I'm going to go with all the bands (and the network will do its job), but as time is a (big) issue, maybe discarding some bands is the way to go…
\n I'm not very helpful here, just letting you know I'm in the same problematic.
Hey, i was checking out human_pixel_masks.npy file and found that it follows the shape (256, 256, 1). May its w.r.t a particular time stamp. Any thoughts on this ?
As mentioned in the data section:
\n\"human_pixel_masks.npy: array with size of H x W x 1 x R. Each example is labeled by R individual human labelers. R is not the same for all samples. The labeled masks have value either 0 or 1 and correspond to the (n_times_before+1)-th image in band_{08-16}.npy. They are available only in the training set.\"
\nIt corresponds to the time stamp (n_times_before+1)
", "rawMarkdown": "As mentioned in the data section: \n\n\"human_pixel_masks.npy: array with size of H x W x 1 x R. Each example is labeled by R individual human labelers. R is not the same for all samples. The labeled masks have value either 0 or 1 and correspond to the (n_times_before+1)-th image in band_{08-16}.npy. They are available only in the training set.\"\n\nIt corresponds to the time stamp (n_times_before+1)", "votes": 1 }, { "id": 2260728, "postDate": "2023-05-15T20:23:23.170Z", "content": "Just realized it when reading the data section again. Thank you so much !!
", "rawMarkdown": "Just realized it when reading the data section again. Thank you so much !!" }, { "id": 2260729, "postDate": "2023-05-15T20:23:46.110Z", "content": "I would say it's an artifact, cause I don't see the difference between (256,256) and (256,256,1) but I might be wrong.
\nConcerning the bands, I read the preprint (quite informative, actually) and they say that some bands are the difference between two other bands. My next move concerning this is to read the .json metadata file, to see if there's something to get from it. (dropping some bands might actually boost the training time)
So, Patchef, this just says that the mask, for all instances, corresponds to the 5th timeframe, am I correct ? (and I think you were mentioning the 'human_individual_mask.npy' file)
", "rawMarkdown": "So, Patchef, this just says that the mask, for all instances, corresponds to the 5th timeframe, am I correct ? (and I think you were mentioning the 'human_individual_mask.npy' file)" }, { "id": 2260736, "postDate": "2023-05-15T20:32:42.630Z", "content": "He is correct. Its the 5th frame a.k.a 4 since counting starts from 0. In n_times_before+1 n_times_before is actually 4. Also, i found this ppt very helpful, do look into it.
Hi Kenni,
\nI'm actually asking myself the same question. In doubt, I'm going to go with all the bands (and the network will do its job), but as time is a (big) issue, maybe discarding some bands is the way to go…
\nI'm not very helpful here, just letting you know I'm in the same problematic.
\n\nHi Kenni,
\n
\n I'm actually asking myself the same question. In doubt, I'm going to go with all the bands (and the network will do its job), but as time is a (big) issue, maybe discarding some bands is the way to go…
\n I'm not very helpful here, just letting you know I'm in the same problematic.
Hey, i was checking out human_pixel_masks.npy file and found that it follows the shape (256, 256, 1). May its w.r.t a particular time stamp. Any thoughts on this ?
As mentioned in the data section:
\n\"human_pixel_masks.npy: array with size of H x W x 1 x R. Each example is labeled by R individual human labelers. R is not the same for all samples. The labeled masks have value either 0 or 1 and correspond to the (n_times_before+1)-th image in band_{08-16}.npy. They are available only in the training set.\"
\nIt corresponds to the time stamp (n_times_before+1)
", "votes": 1, "replies": [] }, { "id": 2260728, "author_name": "Kenni", "author_url": "", "post_date": "2023-05-15T20:23:23.170000", "content": "Just realized it when reading the data section again. Thank you so much !!
", "votes": 0, "replies": [] }, { "id": 2260729, "author_name": "iceman273k", "author_url": "", "post_date": "2023-05-15T20:23:46.110000", "content": "I would say it's an artifact, cause I don't see the difference between (256,256) and (256,256,1) but I might be wrong.
\nConcerning the bands, I read the preprint (quite informative, actually) and they say that some bands are the difference between two other bands. My next move concerning this is to read the .json metadata file, to see if there's something to get from it. (dropping some bands might actually boost the training time)
So, Patchef, this just says that the mask, for all instances, corresponds to the 5th timeframe, am I correct ? (and I think you were mentioning the 'human_individual_mask.npy' file)
", "votes": 0, "replies": [] }, { "id": 2260736, "author_name": "Kenni", "author_url": "", "post_date": "2023-05-15T20:32:42.630000", "content": "He is correct. Its the 5th frame a.k.a 4 since counting starts from 0. In n_times_before+1 n_times_before is actually 4. Also, i found this ppt very helpful, do look into it.