Unable to reproduce the results in the paper

[bush-21]

Hello, I tested some data and could not get the same results as in your paper. Here are two cases I tested. Although the generated results are close to my description, as you said in your paper, this is a new image rather than a continuation of the original image.

case 1

python run_object_sketching.py --num_strokes 16 --num_segments 5 --num_sketches 5 --attention_init 0 --text_prompt "a_simple_drawing_of_a_bicycle" --sketches_edit "input/wheel.png" --output_name "bicycle_w" --train_with_diffusion --control

case 2

python run_object_sketching.py --num_strokes 16 --num_segments 5 --num_sketches 5 --attention_init 0 --text_prompt "a_simple_drawing_of_a_church" --sketches_edit "input/church.png" --output_name "church_w" --train_with_diffusion --control