Open Obikate opened 4 months ago
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!
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?