search

ykdai / BasicPBC

Official Implementation of "Learning Inclusion Matching for Animation Paint Bucket Colorization"
Other
259 stars 23 forks source link

How to prepare my own line drawing? #29

Closed Johanan528 closed 2 months ago

Johanan528 commented 2 months ago

Hi, thanks for your excellent work! We also do a line art colorization work, and would like to cite and use your work as a baseline. How should I prepare my own line drawing for actual testing? Is a 3-channel binary (0/255) line OK? This is my line art.

Thanks again~ 0001 (2) 0000 (2)

ykdai commented 2 months ago

3-channel binary (0/255) line is okay. You also need to generate the paired segment map and json using the generate_paired_data function at https://github.com/ykdai/BasicPBC/blob/main/paint/colorlabel.py. Besides, you have to ensure that your ground truth's colors are discrete for real animation.

Johanan528 commented 2 months ago

Here is my running command, is it correct? Will this automatically generate seg: python inference_line_frames.py --path dataset/test/xx --seg_type trappedball. Additionally, can you explain in more detail the sentence 'Besides, you have to ensure that your ground truth's colors are discrete for real animation.' Thank you very much.

ykdai commented 2 months ago

Sorry for replying late. Since we will calculate the segment-wise accuracy as our metric, if you need to calculate the metrics, the colors of the ground truth should be segment-wise rather than pixel-wise. However, if you just use our method for inference, you can just run the code python inference_line_frames.py --path dataset/test/xx --seg_type trappedball.

Johanan528 commented 2 months ago

Got it, thanks for your reply!