Mask RCNN and RLE

[enemni]

Hello everyone,

I am running the matterport Mask_RCNN on my own dataset created using waspinator/pycococreator

Everything works fine when I am using polygons, but as soon as I am using RLE because I have crowded images I got an error when trying to display a few images from the training dataset,

I got problems with images with holes using polygons and I assumed that is why the pycococreator uses RLE when the image is crowded. This part works perfectly, but then since MaskRCCN requires polygons I am stuck in this loop.

Is there a way to convert RLE to Polygons? I am not even sure this would solve the problem

[waspinator]

you can decode RLE into binary masks using the decode function in the original cocoapi tools, and then use binary_mask_to_polygon to get a polygon.

But the current coco format doesn't encode holes in objects properly, so this may not help solve your issue.

[gethubwy]

Hello everyone,

I am running the matterport Mask_RCNN on my own dataset created using waspinator/pycococreator

Everything works fine when I am using polygons, but as soon as I am using RLE because I have crowded images I got an error when trying to display a few images from the training dataset,

I got problems with images with holes using polygons and I assumed that is why the pycococreator uses RLE when the image is crowded. This part works perfectly, but then since MaskRCCN requires polygons I am stuck in this loop.

Is there a way to convert RLE to Polygons? I am not even sure this would solve the problem

Hello,Excuse me, I want to know how to train a new mode, I use the software named "labelme" to label my data, I want to train a new model starting from pre-trained COCO weights, but I don't know how to write the annotation file in validate set, are there tools to help me? Thank you.

[enemni]

Hello @gethubwy, I am not sure I get your question, can you please share the code with us. labelme should solve your problem and label your entire dataset