TAO DetectNet_v2

[hiteam_yangyan]

Car 0.60 3 -2.42 0.00 185.93 214.05 348.86 1.56 1.57 4.37 -6.96 1.73 7.83 -3.13
Car 0.00 1 -2.73 57.68 178.66 341.72 285.91 1.65 1.68 3.88 -6.88 1.77 12.36 3.05
Car 0.00 2 0.40 133.20 154.23 386.37 263.44 2.08 1.80 4.37 -7.14 1.75 14.97 -0.04
Cyclist 0.00 0 1.66 501.01 178.40 523.53 228.03 1.59 0.58 1.69 -3.29 1.78 24.09 1.53
Car 0.84 0 -0.75 1094.88 190.09 1241.00 374.00 1.54 1.75 3.36 6.04 1.68 5.65 0.04
Car 0.19 3 -0.57 934.82 184.38 1241.00 326.25 1.48 1.52 3.33 5.79 1.63 8.47 0.01
Car 0.00 2 -0.56 882.79 183.25 1122.39 291.95 1.50 1.58 2.81 5.88 1.68 11.09 -0.08
Car 0.00 2 2.68 800.16 181.42 1050.12 273.59 1.57 1.68 3.97 5.77 1.75 13.60 3.08
Car 0.00 2 -0.39 755.13 184.19 967.47 253.54 1.44 1.65 4.26 5.60 1.72 16.48 -0.07
Car 0.00 2 -0.33 749.78 170.88 917.51 243.48 1.77 1.69 3.82 5.72 1.75 18.85 -0.04
Car 0.00 1 -2.90 344.07 180.05 479.10 224.15 1.49 1.59 4.44 -7.12 1.78 26.03 3.12
Car 0.00 2 -2.91 382.70 179.71 495.48 219.10 1.48 1.63 4.10 -6.79 1.78 28.78 -3.14
Car 0.00 0 -1.69 631.71 179.85 678.40 214.37 1.41 1.56 4.08 1.90 1.74 31.98 -1.63
Car 0.00 2 -1.62 619.92 176.90 647.01 199.92 1.43 1.58 3.37 1.47 1.71 47.02 -1.59
Van 0.00 2 -1.58 605.20 168.44 621.13 187.32 2.33 1.96 6.11 0.39 1.83 93.03 -1.57
DontCare -1 -1 -10 521.02 168.77 595.02 193.81 -1 -1 -1 -1000 -1000 -1000 -10

How is this data generated in TAO DetectNet_v2 label? If I want to generate this data myself, how do I do it?

[Morganh]

It is a label file with KITTI format.
Please refer to Data Annotation Format - NVIDIA Docs for more info.

A KITTI format label file is a text file containing one line per object.

For detection the Toolkit only requires the class name and bbox coordinates fields to be populated. This is because the TAO training pipe supports training only for class and bbox coordinates. The remaining fields may be set to 0. Here is a sample file for a custom annotated dataset:

car 0.00 0 0.00 587.01 173.33 614.12 200.12 0.00 0.00 0.00 0.00 0.00 0.00 0.00
cyclist 0.00 0 0.00 665.45 160.00 717.93 217.99 0.00 0.00 0.00 0.00 0.00 0.00 0.00
pedestrian 0.00 0 0.00 423.17 173.67 433.17 224.03 0.00 0.00 0.00 0.00 0.00 0.00 0.00

[hiteam_yangyan]

I customized 10 kitti data format images, why then package into the specified directory, why does it say zip package integrity is missing

[Morganh]

@hiteam_yangyan

why does it say zip package integrity is missing

Could you please share full command and full log? Thanks. I’d like to understand more about your steps and check what is happening.

[hiteam_yangyan]

This may take a while: verify integrity of zip files

!sha256sum $LOCAL_DATA_DIR/data_object_image_2.zip | cut -d ’ ’ -f 1 | grep -xq ‘^351c5a2aa0cd9238b50174a3a62b846bc5855da256b82a196431d60ff8d43617$’ ;
if test $? -eq 0; then echo “images OK”; else echo “images corrupt, redownload!” && rm -f $LOCAL_DATA_DIR/data_object_image_2.zip; fi
!sha256sum $LOCAL_DATA_DIR/data_object_label_2.zip | cut -d ’ ’ -f 1 | grep -xq ‘^4efc76220d867e1c31bb980bbf8cbc02599f02a9cb4350effa98dbb04aaed880$’ ;
if test $? -eq 0; then echo “labels OK”; else echo “labels corrupt, redownload!” && rm -f $LOCAL_DATA_DIR/data_object_label_2.zip; fi

The result is: labels corrupt, redownload!

[hiteam_yangyan]

2023-11-08 01:06:41,011 [TAO Toolkit] [INFO] root 2102: DetectNet V2 model built.
2023-11-08 01:06:41,011 [TAO Toolkit] [INFO] root 2102: Building rasterizer.
2023-11-08 01:06:41,012 [TAO Toolkit] [INFO] root 2102:
Traceback (most recent call last):
File “/usr/local/lib/python3.8/dist-packages/nvidia_tao_tf1/cv/detectnet_v2/scripts/train.py”, line 1067, in
raise e
File “/usr/local/lib/python3.8/dist-packages/nvidia_tao_tf1/cv/detectnet_v2/scripts/train.py”, line 1046, in
main()
File “/usr/local/lib/python3.8/dist-packages/decorator.py”, line 232, in fun
return caller(func, *(extras + args), **kw)
File “/usr/local/lib/python3.8/dist-packages/nvidia_tao_tf1/cv/detectnet_v2/utilities/timer.py”, line 46, in wrapped_fn
return_args = fn(*args, **kwargs)
File “/usr/local/lib/python3.8/dist-packages/nvidia_tao_tf1/cv/detectnet_v2/scripts/train.py”, line 1024, in main
run_experiment(
File “/usr/local/lib/python3.8/dist-packages/nvidia_tao_tf1/cv/detectnet_v2/scripts/train.py”, line 887, in run_experiment
train_gridbox(results_dir, experiment_spec, output_model_file_name, input_model_file_name,
File “/usr/local/lib/python3.8/dist-packages/nvidia_tao_tf1/cv/detectnet_v2/scripts/train.py”, line 682, in train_gridbox
build_rasterizers(experiment_spec,
File “/usr/local/lib/python3.8/dist-packages/nvidia_tao_tf1/cv/detectnet_v2/scripts/train.py”, line 316, in build_rasterizers
bbox_rasterizer = BboxRasterizer(input_width=input_width,
File “/usr/local/lib/python3.8/dist-packages/nvidia_tao_tf1/cv/detectnet_v2/rasterizers/bbox_rasterizer.py”, line 96, in init
assert target_class_name in self.bbox_rasterizer_config
AssertionError
Execution status: FAIL
2023-11-08 09:06:49,594 [TAO Toolkit] [INFO] nvidia_tao_cli.components.docker_handler.docker_handler 337: Stopping container.

This is an error log when I train, I don’t know why

[Morganh]

@hiteam_yangyan

This may take a while: verify integrity of zip files

!sha256sum $LOCAL_DATA_DIR/data_object_image_2.zip | cut -d ’ ’ -f 1 | grep -xq ‘^351c5a2aa0cd9238b50174a3a62b846bc5855da256b82a196431d60ff8d43617$’ ;
if test $? -eq 0; then echo “images OK”; else echo “images corrupt, redownload!” && rm -f $LOCAL_DATA_DIR/data_object_image_2.zip; fi
!sha256sum $LOCAL_DATA_DIR/data_object_label_2.zip | cut -d ’ ’ -f 1 | grep -xq ‘^4efc76220d867e1c31bb980bbf8cbc02599f02a9cb4350effa98dbb04aaed880$’ ;
if test $? -eq 0; then echo “labels OK”; else echo “labels corrupt, redownload!” && rm -f $LOCAL_DATA_DIR/data_object_label_2.zip; fi

The result is: labels corrupt, redownload!

Can you download data_object_label_2.zip again?
Please refer to https://github.com/NVIDIA/tao_tutorials/blob/main/notebooks/tao_launcher_starter_kit/detectnet_v2/detectnet_v2.ipynb

Seems that the file you downloaded is corrupted.