提交 75a63c4e 编写于 作者: W wuzewu

Update detection demo

上级 e1cd9d3d
......@@ -14,7 +14,7 @@ parser = argparse.ArgumentParser(__doc__)
parser.add_argument("--num_epoch", type=int, default=50, help="Number of epoches for fine-tuning.")
parser.add_argument("--use_gpu", type=ast.literal_eval, default=True, help="Whether use GPU for fine-tuning.")
parser.add_argument("--checkpoint_dir", type=str, default="faster_rcnn_finetune_ckpt", help="Path to save log data.")
parser.add_argument("--batch_size", type=int, default=1, help="Total examples' number in batch for training.")
parser.add_argument("--batch_size", type=int, default=5, help="Total examples' number in batch for training.")
parser.add_argument("--module", type=str, default="faster_rcnn_resnet50_coco2017", help="Module used as feature extractor.")
parser.add_argument("--use_data_parallel", type=ast.literal_eval, default=False, help="Whether use data parallel.")
# yapf: enable.
......@@ -54,7 +54,7 @@ def finetune(args):
config = hub.RunConfig(
log_interval=10,
eval_interval=10,
eval_interval=100,
use_data_parallel=args.use_data_parallel,
use_pyreader=True,
use_cuda=args.use_gpu,
......
Markdown is supported
0% .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑!
想要评论请 注册