tensorflow_object_detection_create_coco_tfrecord

tensorflow_object_detection_create_coco_tfrecord copied to clipboard

Need to read the images one by one and write as TFRecord, instead of loading all images in memory and create a TF record

2

Akshai

ImportError: No module named 'dataset_util'

1

Hi, which "dataset util" do you use? I cannot find the proper one. Thank you!

shi27feng

Hi, It seems that you get the Object detection API working on https://github.com/tensorflow/models/tree/master/research/object_detection.. Could you please let me know what version of tensorflow and cudnn were you using to make...

lyltencent

Parameters

3

Hey, just a comment, could you change the input parameters. You say: python create_coco_tf_record.py --data_dir=/path/to/your/coco/root/directory \ --set=train \ --output_path=/where/you/want/to/save/pascal.record --shuffle_imgs=True but it should be: python create_coco_tf_record.py --data_dir=/path/to/your/coco/root/directory \ --set=train \...

vertcli

Why do you set an upper limit on the number of images to be read?

1

https://github.com/MetaPeak/tensorflow_object_detection_create_coco_tfrecord/blob/master/create_coco_tf_record.py#L64

legokichi

added a loop of loading image to convert all images

bit1002lst