Hi it is Peter from the video. This is a bit longer conversation but I'm happy to have it here or here: github.com/roboflow/notebooks/discussions/categories/q-a . What would be normalization technique that you'd like to use look like?
Great video, congratulations! I've been using Detetron2 for about a year and I'm really enjoying it. I created my own model and I'm developing a dataset for my purposes. The only thing I couldn't do was customize the colors of the instances, I need each group to have its own color (as if it were a legend) so I can better visualize the project I'm developing. Do I have any function or library that could do this type of customization? Thanks! Bests from Brazil
Hi! Do you mean tutorial on training some vision transformers using Detectron2? If you have some ideas for tutorials that we should do, please submit a request in our Notebooks repository on GitHub: github.com/roboflow-ai/notebooks
cfg.MODEL.MASK_ON = False - If your dataset does not contain segmentation masks, and you only need to detect bounding boxes, change the model configuration to perform object detection instead of instance segmentation.
CUDA error: device-side assert triggered CUDA kernel errors might be asynchronously reported at some other API call, so the stacktrace below might be incorrect. For debugging consider passing CUDA_LAUNCH_BLOCKING=1. Compile with `TORCH_USE_CUDA_DSA` to enable device-side assertions. i got error like this, what should i do sir?thank you
@@Roboflow can detectronV2 detect multi objects? I ran the code on my dataset which has 3 class, after training, the results show that 2 classes detect and segment fine with at least 90 percent accuracy but the third class did not detect at all!
When I use this code # Look at training curves in tensorboard: %load_ext tensorboard %tensorboard --logdir $OUTPUT_DIR_PATH It got error 403. How can I fix it?
Thanks for great video I just have one doubt, don't we need to normalize the custom data and its bounding box before using it?
Hi it is Peter from the video. This is a bit longer conversation but I'm happy to have it here or here: github.com/roboflow/notebooks/discussions/categories/q-a . What would be normalization technique that you'd like to use look like?
Nice video!! Thanks for sharing!
thank you very much! ;)
Great video, congratulations! I've been using Detetron2 for about a year and I'm really enjoying it. I created my own model and I'm developing a dataset for my purposes. The only thing I couldn't do was customize the colors of the instances, I need each group to have its own color (as if it were a legend) so I can better visualize the project I'm developing. Do I have any function or library that could do this type of customization? Thanks! Bests from Brazil
Can I use Object Detection(Bounding Box) instead of Instance Segmentation for Detectron2 ?
please make video about training detectron2 projects like VITDet
Hi! Do you mean tutorial on training some vision transformers using Detectron2? If you have some ideas for tutorials that we should do, please submit a request in our Notebooks repository on GitHub: github.com/roboflow-ai/notebooks
Why would you use Detectron2 over YoloV8 or Mask_RCNN? - I'm still a bit confused about the advantages of Detectron2 over the rest.
did u find an answer to this?
@@alocatnaf no :(
asking the same question
Hi can you please tell me how to use this trained model to detect the objects from fresh images?
I am getting error ;
AttributeError: Cannot find field 'gt_masks' in the given Instances!
can you please suggest how to solve it .
Hi could you please create a bug report here: github.com/roboflow/notebooks I'll try to help you out :)
cfg.MODEL.MASK_ON = False - If your dataset does not contain segmentation masks, and you only need to detect bounding boxes, change the model configuration to perform object detection instead of instance segmentation.
Is there any change of code while my data just have 1 class? I can't visualize the predict mask btw
how can we get approx size of the object? or is there any way to do it? can it be done while labelling the images if yes then how?
CUDA error: device-side assert triggered
CUDA kernel errors might be asynchronously reported at some other API call, so the stacktrace below might be incorrect.
For debugging consider passing CUDA_LAUNCH_BLOCKING=1.
Compile with `TORCH_USE_CUDA_DSA` to enable device-side assertions.
i got error like this, what should i do sir?thank you
did you solve it ?
make sure your NUM_CLASSES are the same as your dataset's
@@abdallahmashaly136 make sure your NUM_CLASSES are the same as your dataset's
Hi, great video, and works for me, generally which one is better for object detectron, Detectronv2 or YOLOv8?
Do you care more about the speed or the accuracy?
@@Roboflow in the case I am working now, the accuracy is more important
@@lofihavensongs Id go with Detectron2 ;)
@@Roboflow can detectronV2 detect multi objects? I ran the code on my dataset which has 3 class, after training, the results show that 2 classes detect and segment fine with at least 90 percent accuracy but the third class did not detect at all!
@@lofihavensongs yes it can. What is the class distribution in your dataset?
Hi can anyone tell me how to chose which yaml file is best suitable for our model?
When I use this code
# Look at training curves in tensorboard:
%load_ext tensorboard
%tensorboard --logdir $OUTPUT_DIR_PATH
It got error 403.
How can I fix it?
I would like to know how to create a Confusion matrix. please.
did u find an answer to this?