Object Detection toolkit based on PaddlePaddle. It supports object detection, instance segmentation, multiple object tracking and real-time multi-person keypoint detection.
Apache License 2.0
12.81k
stars
2.89k
forks
source link
Getting error while to train mask_rcnn for custom coco dataset. #4267
I am trying to train my custom coco data set for instance segmentation using mask rcnn but I got the error AssertionError: not found any coco record in dataset/COCO/annotations/train.json. For using my custom dataset, namely COCO, I modified the coco_instance.yml and provided the path of my custom datasets (image + annotation). Here is the snippet of the path modification
here are the error messages while I try to train the mask rcnn model.
to get rid of this error, I modified again my custom dataset path in the file coco_instance.yml from / to \ after that I got the another error ValueError: Dataset dataset\COCO is not valid and cannot parse dataset type 'dataset\COCO' for automaticly downloading, whichonly supports 'voc' , 'coco', 'wider_face', 'fruit', 'roadsign_voc' and 'mot' currently Here are the snippet of the error and modification.
I found the same issue but for language barrier I could not solve my problem by getting ideas from the previous issue.
This problem raise for providing wrong class number in the file coco_instance.yml. My data set had 2 classes, I for got give number of classes 2+1 (1 for background).
I am trying to train my custom coco data set for instance segmentation using
mask rcnn
but I got the errorAssertionError: not found any coco record in dataset/COCO/annotations/train.json
. For using my custom dataset, namely COCO, I modified thecoco_instance.yml
and provided the path of my custom datasets (image + annotation). Here is the snippet of the path modification here are the error messages while I try to train the mask rcnn model. to get rid of this error, I modified again my custom dataset path in the filecoco_instance.yml
from/
to\
after that I got the another errorValueError: Dataset dataset\COCO is not valid and cannot parse dataset type 'dataset\COCO' for automaticly downloading, whichonly supports 'voc' , 'coco', 'wider_face', 'fruit', 'roadsign_voc' and 'mot' currently
Here are the snippet of the error and modification. I found the same issue but for language barrier I could not solve my problemby
getting ideas from the previous issue.