I haven't seen anyone else have this issue, which leads me to believe it is a trivial problem that should hopefully be not that hard to fix. I am not the strongest coder, which is also what may have led me to find myself in this predicament. There are three long screenshots of the errors, but hopefully this post isn't too long.
I am having issues executing the crop_dataset.py code using my own dataset. I followed the instructions of having my images be .jpg, I don't have any labels, and I also changed the train_config.yml as instructed. That is all the editing that I saw to do on the "Bringing your own data" section of the Github. When I first run the crop_dataset.py after changing only the config file, I this error:
I believe the top half of the error is just warnings, but the bottom half says it tries to open cityscapes.py, which isn't what I'm trying to use. I then go into the crop_dataset.py file and change the "dataset_name" at line 140 to my dataset name, which is "TS_dataset". I then get this error that I've gotten before of an unknown dataset (I also tried using the exact name "dataset_name" but got the same issue):
So I go look in data.py at the specified line and see that line is looking for a dataset called "directory", so in desperation I try to rename my dataset "directory", which I'm fairly sure isn't going to help, but at least I get a different error:
My guesses as to why these errors are occurring are that maybe I accidentally have a package in my STEGO environment updated to a newer version that is causing issues, or perhaps I am not formatting the data correctly, or even I have not edited all of the file paths or other files correctly/completely. But I am not able to really figure out what the exact issue is. Any help would be greatly appreciated.
Thank you very much,
Bobby
PS: this is what my data look like. They are jpgs and I added the ".jpg" extension to the files later on. With or without that ".jpg" I continued getting the "Unknown dataset" error every time I ran the crap_datasets.py. They are all 3300x2100 pixels in size as well.
Hello,
I haven't seen anyone else have this issue, which leads me to believe it is a trivial problem that should hopefully be not that hard to fix. I am not the strongest coder, which is also what may have led me to find myself in this predicament. There are three long screenshots of the errors, but hopefully this post isn't too long.
I am having issues executing the crop_dataset.py code using my own dataset. I followed the instructions of having my images be .jpg, I don't have any labels, and I also changed the train_config.yml as instructed. That is all the editing that I saw to do on the "Bringing your own data" section of the Github. When I first run the crop_dataset.py after changing only the config file, I this error:
I believe the top half of the error is just warnings, but the bottom half says it tries to open cityscapes.py, which isn't what I'm trying to use. I then go into the crop_dataset.py file and change the "dataset_name" at line 140 to my dataset name, which is "TS_dataset". I then get this error that I've gotten before of an unknown dataset (I also tried using the exact name "dataset_name" but got the same issue):
So I go look in data.py at the specified line and see that line is looking for a dataset called "directory", so in desperation I try to rename my dataset "directory", which I'm fairly sure isn't going to help, but at least I get a different error:
My guesses as to why these errors are occurring are that maybe I accidentally have a package in my STEGO environment updated to a newer version that is causing issues, or perhaps I am not formatting the data correctly, or even I have not edited all of the file paths or other files correctly/completely. But I am not able to really figure out what the exact issue is. Any help would be greatly appreciated.
Thank you very much, Bobby
PS: this is what my data look like. They are jpgs and I added the ".jpg" extension to the files later on. With or without that ".jpg" I continued getting the "Unknown dataset" error every time I ran the crap_datasets.py. They are all 3300x2100 pixels in size as well.