Відео

i want to train model on custom data how can i do that ?

4:13 media offline

paddler or mmocr which is most accurate?

Could you provide the code? The GitHub link is not working.

This solution is same as commenting the whole code to avoid error instead of finding the bug.

This solution is same as commenting the whole code to avoid errors instead of finding a bug.

Sir, Would you mind giiving me this solution of this porblem . when I want to run your code in terms of solar panel image then it is creating this problem like "SupervisionWarnings: green is deprecated: `Color.green()` is deprecated and will be removed in `supervision-0.22.0`. Use `Color.GREEN` instead. SupervisionWarnings: annotate is deprecated: `BoxAnnotator` is deprecated and will be removed in `supervision-0.22.0`. Use `BoundingBoxAnnotator` and `LabelAnnotator` instead" . Original code of your is "import supervision as sv import numpy as np mask_annotator = sv.MaskAnnotator() box_annotator = sv.BoxAnnotator(color=sv.Color.green()) detections=sv.Detections.from_sam(result) annotated_image = box_annotator.annotate(scene=image_bgr.copy(),detections=detections,skip_label=True) annotated_image = mask_annotator.annotate(scene=annotated_image.copy(),detections=detections) sv.plot_images_grid( images=[image_bgr,annotaed_image], grid_size=(1,2), titles=['Original_image','Annotated_image'], )"

@siliconvision does the notebook still work?

How to run with gpu instead of cpu?

How can we perform it on whole dataset with multi class classification? Do you have any notebook for it?

I can't find the same path 😢

doesthis work for text too as in copy original signature to paste in a document?

Hi, I tried to implement on Raspberry Pi but has the "too many value to unpack" error. Any advice? I have trimmed from 5 parameters to 4, and change the first param to "boxes.

Can share the python codes?

I have been looking for something this. Excellent quality. Loved that you also talked about multi-lingual support! Thank you! 🙂

Hi, I tryed to do the same method but I get a very incorrectly colored generated images, then i tryed SDXL instead of SD v2 and now I get black generated images, i would appreciate if you know how to solve this.

do we need to set the cuda_home first?

Its great! I am recently working on character, face, object recognition. Suggest the best libraries for: 1. Face (I use Opencv LBPH..) 2. Image to text? 3. Table Image to csv 4. Objects recognition? You should make more similar videos, soon you will be monetized.

Could you please provide source code

could you please provide this colab notebook

colab is = a stable diffision????? I didn know that

Thanks i really wanted this 😊

👏👏👏

Wow nice video!

Incredibly high-quality overview! Thank you!

Nice and easy way of programming. Thanks for showing all the steps!

do you freelance ? i need to segment solar panels from UAV images. If yes, can you share your email id ?

thank you

thank you

excellent 🙂🙂

did not solve the problem.

Hi could you please help us with our project? We want to detect all text on screenshot of software. This is a more complex task than we tough, Tesseract and EasyOCR doesn't work. We did a lot of preprocessing (binarization, denoise, scale*3, grayscale...) and managed to go up to 90% detection on tesseract on most software. Our software is an AI that answer question based on what it saw on the screen ( It is called onistep ). This is a desktop app. This will be used by all our client on their end so we cannot use the GPU and the answer must take 5 second maximum. We think that solution like MMocr, keras OCR or paddle OCR could work but we never worked with machine learning based OCR. Do you think you can work with us on this? Do you think it is doable? Of course it would be paid work. I also sent you an email with some image preprocessed if you wanna test our image.

Nice video brom, it helped me a lot.

I think your notebook is a bit outdated because groundino module is not working. missing some installation

this doesn't solve the problem at all...... this switches Torch to run on your CPU instead of GPU, so it will be slow as hell

Can you do about face detection and face regconition?

Great tutorial sir but while solving I get "size mismatch for wrapped_model.backbone.layer1.0.downsample.1.weight: copying a param with shape torch.Size([256]) from checkpoint, the shape in current model is torch.Size([256, 64, 1, 1])." this error. how can I solve it ?

could you share google colab link for your notebook?

that is not solving anything what about pepole who actually want to use cuda??

Solve? or Avoid the problem.

"promo sm" 😱

If I only want to extract 1 object from an image, it doesn't work. For example if I want to detect a bear, i have to set the classes = ['bear', 'bear'] and get 2 extractions of the same bear. Also, when I print detection class_ids, it says None, None instead of 0,1. I have to manually set class_id = [1, 0]. Also, how can I download the final extracted objects instead of just plotting them?

Brother ho can i contact you

Brother how i can i contact you