How to use Augmentation in TLT Python?

Hey I m trying to use rotation spatial augmentation in the TLT training pipeline but It is not working

spatial_augmentation {
hflip_probability: 0.5
vflip_probability: 0.0
zoom_min: 1.0
zoom_max: 1.0
translate_max_x: 8.0
translate_max_y: 8.0
rotation_angle_max: 30.0
rotation_units: "degrees"


Error: google.protobuf.text_format.ParseError: 48:5 : Message type “AugmentationConfig.SpatialAugmentation” has no field named “rotation_angle_max”.

I tried the way mentioned here in the documentation which tells us to use it like this

angle: 5.0
units: "degrees"

But it is also not working, and I see that the code is not open source.
So please tell me the right FIELD name which can be used for using the rotation Augmentation in the detection pipeline.



Please note that it is only compatible with the tool augment. See Offline Data Augmentation — Transfer Learning Toolkit 3.0 documentation

In order to circumvent this and generate a dataset with the required augmentations and give control to the user, TLT provides an offline augmentation tool called augment .

Please download latest jupyter notebook.

wget --content-disposition -O
unzip -u -d ./tlt_cv_samples_v1.0.2 && rm -rf && cd ./tlt_cv_samples_v1.0.2

then, go into tlt_cv_samples_v1.0.2/augment.
There is augment.ipynb and its specs folder.

# Spec file for tlt-augment.
angle: 10.0
units: “degrees”
image_path: “image_2”
label_path: “label_2”
hue_rotation_angle: 5.0
saturation_shift: 1.0
output_image_width: 1248
output_image_height: 384
output_image_channel: 3
image_extension: “.png”