content_aware task，own dataset，different image size

[cyj95]

I wonder what parameters need to change when I have different image size? my image size is 1024*1024， content_aware task in utils.py I modify "posterlayout": (102, 150) to (1024, 1024) in visualization section, I only want to save the first image. from visualization import ContentAwareVisualizer, create_image_grid visualizer = ContentAwareVisualizer() images = visualizer(ranked_response, processed_test_data[test_idx]["idx"]) images[0].save("images0.png","PNG")

but there is something wrong when I check the layout box. some boxes are drawn out the poster. how should I fix this?

[kaoshizhou]

Hi @cyj95,

Sorry for the late response. Actually, we can not reproduce the issue you mentioned. Below is a generated layout with the canvas size of 1024 * 1024, and the boxes are all within the canvas.

Make sure you have reconfigured the canvas size in utils.py before processing the raw data. Otherwise, you should remove the processed data and process raw data again. Feel free to contact us if you have any further questions.

[ruick]

Hi Thank you so much for the amazing work. Could you help me with how can LayoutPrompter used on my own data to generate layout. it seems I have to use the data input such as train or test data. Thank you!