I am attempting to train a NN to make predictions within the SA platform. The annotations are all rotated ellipses (i.e., object detection), 2 classes. I am getting an error that simply states that "out of 66 test images 66 had unusable annotations", but it provides no additional details as to how to identify the offending annotations or how to remedy the situation. Any help would be very much appreciated!
Posted by Brandon Sackmann 29 days ago
I'm following the OpenCV AI kit tutorial and one of the topics is about model deployment in OAK. I got stuck at step#3 of "SuperAnnotate_OAK_YOLOv4_tiny_Deployment" in COLAB. It returns error of missing "classes.json" file like below. Can somebody help on this? < FileNotFoundError Traceback (most recent call last) <ipython-input-15-226e3c0ca04b> in <module>() 23 class_names = set() 24 for folder_name in proj_folder.values(): ---> 25 with open('/content/training_data/' + folder_name +'/classes/classes.json', 'r') as f: 26 proj_class_data = json.load(f) 27 for class_entry in proj_class_data: FileNotFoundError: [Errno 2] No such file or directory: '/content/training_data/proj_0/classes/classes.json' />
Posted by JuanYi about a month ago
Hello, I'd like to make cuboid annotations with multiple vanishing points (the current tool allows only for one). This can be actually as simple as unlocking the cuboid vertices positioning. Any idea if this feature will be added soon? Many thanks!
Posted by Alessandro Melis about a month ago
Simple question, can I create a custom model using my partially annotated data to help me run smart prediction and automatically annotate the rest? So far I see I can use pre-defined models but I hope I can use my custom model instead.
Posted by Carlos Argueta about a month ago
How do I create a bulk export of data based on parameters found on the UI such as "fused images" and status like "completed". I've found this one method: sa.prepare_export( project = "Lane Annotation", folder_names = ["Batch1", "Batch2"])
Posted by adelle 2 months ago
I am trying to use the SDK to convert annotations from VOC or COCO to your format, but am having some trouble. Your documentation does not specify what folder structure your code expects, and keeps reporting that it is not finding what it is looking for: SA-PYTHON-SDK - INFO - All files with following extensions ['jpg', 'jpeg', 'png', 'tif', 'tiff', 'webp', 'bmp'] will be copied to output folder SA-PYTHON-SDK - WARNING - Images doesn't exist SA-PYTHON-SDK - WARNING - 'Annotations' directory is empty SA-PYTHON-SDK - INFO - Converting to SuperAnnotate JSON format 0it [00:00, ?it/s] SA-PYTHON-SDK - INFO - Conversion completed I have tried with the standard VOC and COCO datasets, and still get the same error. If you could specify what folder structure you expect, it would be very useful.
Posted by Angela King 2 months ago
Hi, I wonder if there is a example project or template for the OAK-D colab example? https://colab.research.google.com/github/superannotateai/model-deployment-tutorials/blob/main/OAK/SuperAnnotate_OAK_Deeplabv3%2B_Deployment.ipynb#scrollTo=GhvXpjlY-G3e Kind regards Julian
Posted by Julian 2 months ago