Tao training format valiation

Accelerated Computing Intelligent Video Analytics TAO Toolkit
1

When executing tao training, is it possible to specify the format?

I would like to use yolo or coco format.

2

Do you mean the format of label file?

3

Yes d. Yes.
I would like to do TRAINING with the TEXT of LABEL in YOLO format or COCO or other format.

4

Refer to Data Annotation Format — TAO Toolkit 3.22.05 documentation
Since TAO Toolkit 3.0-22.05, all object detection models support COCO format.

5

but, it will occur python module error.
how can I avoid this error ?
is there docker images for creating coco tf records?

■execute command

python create_coco_tf_record.py
  --logtostderr
  --train_image_dir=$TRAIN_IMAGE_DIR
  --val_image_dir=$VAL_IMAGE_DIR
  --train_object_annotations_file=$TRAIN_COCO_ANNOTATION_FILE
  --val_object_annotations_file=$VAL_ANNOTATION_FILE
  --output_dir=$OUTPUT_DIR

"ModuleNotFoundError: No module named 'research'"

■python3 -m install research

  note: This error originates from a subprocess, and is likely not a problem with pip.
  ERROR: Failed building wheel for pandas
  Running setup.py clean for pandas
Successfully built research
Failed to build pandas
Installing collected packages: pandas, research
6

As shown below, should I to execute to run create_coco_tf_record.py ?

■maskrcnn.ipynb

!tao mask_rcnn run bash $SPECS_DIR/download_and_preprocess_coco.sh $DATA_DOWNLOAD_DIR

this docker image is using python3.6.
so, I must to do that in docker image container. correct ?

(launcher) ubuntu@ip-172-31-0-107:/tmp$ docker images
REPOSITORY                          TAG                     IMAGE ID       CREATED       SIZE
nvcr.io/nvidia/tao/tao-toolkit-tf   v3.22.05-tf1.15.4-py3   ca92a571a959   3 weeks ago   16.1GB
7

You can run it inside the tao docker.
$ tao mask_rcnn run /bin/bash

then,
# python3 xxx

8

sorry.
is this correct ?

(launcher) ubuntu@ip-172-31-0-107:~$ tao mask_rcnn run /bin/bash
2022-06-14 04:18:31,320 [INFO] root: Registry: ['nvcr.io']
2022-06-14 04:18:31,417 [INFO] tlt.components.instance_handler.local_instance: Running command in container: nvcr.io/nvidia/tao/tao-toolkit-tf:v3.22.05-tf1.15.5-py3
2022-06-14 04:18:31,428 [INFO] tlt.components.docker_handler.docker_handler: The required docker doesn't exist locally/the manifest has changed. Pulling a new docker.
2022-06-14 04:18:31,428 [INFO] tlt.components.docker_handler.docker_handler: Pulling the required container. This may take several minutes if you're doing this for the first time. Please wait here.
...
Pulling from repository: nvcr.io/nvidia/tao/tao-toolkit-tf

below, I tried. but this error occur.

(launcher) ubuntu@ip-172-31-0-107:~/development$ docker images
REPOSITORY                          TAG                     IMAGE ID       CREATED       SIZE
nvcr.io/nvidia/tao/tao-toolkit-tf   v3.22.05-tf1.15.4-py3   ca92a571a959   3 weeks ago   16.1GB
↓
docker run -it -v /home/ubuntu/development:/home/ubuntu/development ca92a571a959 /bin/bash
↓
root@82c58989ad2b:/home/ubuntu/development# tao info
bash: tao: command not found
9

I need run docker and derectory volume mount.
so, I want to ask mount methods.

so, how should I do to volume mount（image directory） and docker run.

root@8274918f1fc8:/workspace# ls -la
total 288
drwxr-xr-x 1 root root   4096 May 22 00:23 .
drwxr-xr-x 1 root root   4096 Jun 14 04:25 ..
-rw-r--r-- 1 root root 135165 May 22 00:03 EULA.pdf
-rw-rw-r-- 1 root root 144823 Feb  9 18:55 NVIDIA_Deep_Learning_Container_License.pdf
-rw-r--r-- 1 root root   1261 May 22 00:03 README.md
root@8274918f1fc8:/workspace#

■this ec2 terminal

(launcher) ubuntu@ip-172-31-0-107:~/development/workspace/cat/coco/Cats_5/dataDownloadDir$ ll
total 28
drwx------ 5 ubuntu ubuntu 4096 Jun 14 02:04 ./
drwx------ 6 ubuntu ubuntu 4096 Jun 14 02:00 ../
-rw------- 1 ubuntu ubuntu  125 Jun 14 01:47 README.dataset.txt
-rw------- 1 ubuntu ubuntu  394 Jun 14 01:47 README.roboflow.txt
drwx------ 2 ubuntu ubuntu 4096 Jun 14 01:47 test/
drwx------ 2 ubuntu ubuntu 4096 Jun 14 02:06 train/
drwx------ 2 ubuntu ubuntu 4096 Jun 14 01:47 valid/
10

now, this shell commad works.

root@d798ad4d9879:/home/ubuntu/development/cv_samples_v1.2.0/mask_rcnn/specs# ll
total 44
drwx------ 2 1000 1000  4096 Jun 11 00:59 ./
drwx------ 4 1000 1000  4096 Jun 14 03:04 ../
-rw------- 1 1000 1000   692 Aug 24  2021 coco_labels.txt
-rw------- 1 1000 1000 12311 Aug 24  2021 create_coco_tf_record.py
-rw------- 1 1000 1000  4347 Aug 24  2021 download_and_preprocess_coco.sh
-rw------- 1 1000 1000  2001 Aug 24  2021 maskrcnn_retrain_resnet50.txt
-rw------- 1 1000 1000  2039 Aug 24  2021 maskrcnn_train_resnet50.txt
root@d798ad4d9879:/home/ubuntu/development/cv_samples_v1.2.0/mask_rcnn/specs# bash download_and_preprocess_coco.sh /home/ubuntu/development/workspace/cat/coco/Cats_5/dataDownloadDir
11

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.