Properties in Excel file & Yolov5x6 datasets.

[Wingvisible]

Hi, I have 2 questions regarding the metadata file. I want to use Excel for my metadata structure: how do I put the properties in it? (Do I have to put it like that: hat:cap accessory:necklace ?)

And can I put, in only one Excel file, multiple nfts with different properties for each?

[maximedrn]

Like this:

[["Type", "Name"], ["Type", "Name"], ["Type", "Name"], ["Type", "Name"], ...]

[maximedrn]

You can create your own dataset for these objects, then you have to implement it in the recaptcha.py file.

[Wingvisible]

Hum, I'm not very good with programming so I'm not sure how I'm supposed to do this. I downloaded some datasets on roboflow but I don't know how I'm supposed to implement them in the recaptcha.py file.

[maximedrn]

It requires a Yolov5-compatible dataset, and then you need to train that dataset to generate a pre-trained .pt file.

Documentation:

• Yolov5x6: https://github.com/ultralytics/yolov5
• Train custom data on Yolov5x6: https://github.com/ultralytics/yolov5/wiki/Train-Custom-Data

[Wingvisible]

Ayt! I have my pre-trained .pt file. What do I do next? Thank you for your time!

[maximedrn]

First you have to make sure that the name of the object used in the .yml file is in lower case (crosswalk).
Then place your file in the bot's directory (for example in the yolov5/ folder).
Finally, you will have to edit the recaptcha.py file:

• Lines 66 to 74:

   def load_models(self) -> None:
       """Load Real-ESRGAN, Yolov5's basic and custom models."""
       self.download_models()
       self.upsampler = RealESRGANer(scale=3, model=RRDBNet(
           scale=4, num_in_ch=3, num_out_ch=3, num_feat=64, num_block=23,
           num_grow_ch=32), model_path=abspath(self.model_path[0]))
       self.yolo = load(abspath('yolov5'), 'custom', path=abspath(
           self.model_path[1]), source='local')
      # Add your `.pt` file here:
      self.crosswalks = load(abspath('yolov5'), 'custom', path=abspath(
           'yolov5/crosswalks.pt'), source='local')
      self.specific_classes = ['crosswalks']  # Add the name of the object here.