Share checkpoints and scripts to run the trained model

[Obikate]

I would really like to apply your model on PCB images, but I can see that the necessary checkpoints and scripts for that are not in the repository. Could you please add them?

[betogulliver]

I've successfully trained the model using the FPIC dataset (loss curve LGTM)

could you please share a simple script to do inference for a single image?

I've tried doing the following:

pred = model.predict(im_array)

the shape/dtype of the pred is
(1, 512, 512, 26) float32

512, 512 = im_size
26 = num_classes + bg

are these 'pred' results the logits of the prediction?
how can I convert them into a colored mask images?

thanks again sharing your work, much appreciated!

[PladsElsker]

@betogulliver Do you still have the dataset? Can you share it on hugging face if the license allows it? The FPIC link provided in the repo is dead and the authors are unresponsive.

Edit:
Nevermind found it here
https://physicaldb.ece.ufl.edu/index.php/fics-pcb-image-collection-fpic/

Edit 2:
Nevermind nevermind, there are only 271 images in the dataset I linked, so it seems to only be a sample.