Deep View Validator Supported Datasets
This article will show the dataset formats that is currently supported in Deep View Validator.
Dataset Formats
Currently validation in the EVK (ARM architectures) only supports Darknet datasets because TFRecord datasets consume too much memory and poses TensorFlow issues that still needs to be closely inspected. Typical machines with x86 architectures can support both Darknet and TFRecord datasets.
Deep View Validator accepts a path to a dataset that points to a YAML file or a directory that contains the validation dataset.
- For TFRecord datasets, the directory should contain *.tfrecord files or the YAML file should be structured in the following way if relying on the
--local_reader
parameter in validator.
classes:
- class_1
- class_2
- class_3
validation:
path: <path to the tfrecord files>
- For Darknet datasets, the directory should contain images (jpg, png, jpeg, JPG, PNG, JPEG) and annotations (txt or JSON). The following directory structure is also accepted.
Directory Path
|---------images
|---validate
|------- image.jpg ....
|---------labels
|---validate
|------- image.txt ....
|---------labels.txt (optional)
The labels.txt is an optional file which contains unique string labels in the dataset. Models will output integer indices denoting the label. Deep View Validator will use the labels.txt file provided to convert the integers into their string representations. For example, if labels.txt contains:
ace
two
king
At index 0 would be 'ace', at index 1 would be 'two' and at index 3 would be 'king'.
The YAML file should be structured in the following way if relying on the --local_reader
parameter in validator.
classes:
- class_1
- class_2
- class_3
type: darknet
validation:
annotations: <path to the annotation text files>
images: <path to the image files>
- For more information on 2D detection annotation files.
- For more information on 3D detection annotation files.
- For more information on segmentation annotation files.
- For more information on head pose annotation files.
Note: segmentation and pose annotation files are stored in JSON files, whereas detection annotation files are commonly stored in text files. An exception would be 3D detection which is stored in JSON files as well.
Conclusion
This article has shown how to format the datasets that is supported in Deep View Validator. This includes either Darknet or TFRecord datasets which validator accepts a path to the directory or a path to the YAML file that contains the information of the dataset.
Comments
0 comments
Please sign in to leave a comment.