Config for evaluating a custom dataset

[edwardnguyen1705]

Dear @WXinlong ,

After training on a custom dataset (inherited from the CocoDataset, just like the way in the cityscapes.py), the visualization looks great, but mAP = 0. I see in the test_insts.py coco_eval(result_file, eval_types, dataset.coco)

Is it the right way?

[lucasjinreal]

I think dataset.coco means the object of COCO. which is your loaded coco format dataset. it can be considerate as ground truth coco for evaluation.

[lucasjinreal]

for CustomDataset evaluation got mAP=0, check the annotation iscrowd enable or not, if you using RLE for representation of segmentation mask, you should enable iscrowd=1 when calculate mask iou

[edwardnguyen1705]

@jinfagang Thanks for sharing.

for CustomDataset evaluation got mAP=0, check the annotation iscrowd enable or not, if you using RLE for representation of segmentation mask, you should enable iscrowd=1 when calculate mask iou

I do not quite understand this one. Yes, my dataset uses RLE for representation of segmentation mask.

[lucasjinreal]

@edwardnguyen1705 coco loadRes will force iscrowd to false. You can take a close look at coco evaluation code.

[edwardnguyen1705]

@jinfagang , Thank you for your response.

[haibucuo0]

@edwardnguyen1705 ,how to modify test_ins.py when I use own data？thanks

[edwardnguyen1705]

@haibucuo0 you can take a look at my fork: solo