Closed abdo-ameen closed 3 years ago
Currently Pascal VOC and COCO dataset formats are supported. You can dump your custom dataset into either of these two formats (e.g. in cvat) and just point the train.py
to it, specifying format with --dataset-style
parameter, path to it with --dataset
parameter and image set for VOC or annotation filename for COCO for training and validation stages with --train-image-set
and --val-image-set
keys (they are set with train
and val
by default, correspondingly).
thank you so much
how can I train this model on my own custom dataset?