# PoseClassificationNet

PoseClassificationNet takes a sequence of skeletons (body poses) as network input and predicts the actions of one or more persons in those frames. The model supported in the current version is based on the spatial-temporal graph convolutional network (ST-GCN), which is the most commonly used baseline for skeleton-based action recognition due to its simplicity and computational efficiency. Unlike pixel-based action recognition, ST-GCN is able to exploit the local pattern and correlation from a spatial-temporal graph of human skeletons. This model can be used to train graph convolutional networks (GCNs) for other purposes through transfer learning. Newer architectures with state-of-the-art performance will be released in the future. TAO Toolkit provides the network backbone for 3D poses.

## Preparing the Dataset

PoseClassificationNet requires a sequence of skeletons (body poses) for input. The coordinates need to be normalized. For example, 3D joints are produced relative to the root keypoint (i.e. pelvis) and normalized by the focal length (1200.0 for 1080P). The entrypoint for dataset conversion generates an array of spatio-temporal sequences based on the output JSON metadata from the deepstream-bodypose-3d app.

The input data for training or inference are formatted as a NumPy array in five dimensions (N, C, T, V, M):

• N: The number of sequences

• C: The number of input channels, which is set to 3 in the NGC model

• T: The maximum sequence length in frames, which is 300 (10 seconds for 30 FPS) in the NGC model

• V: The number of joint points, set to 34 for the NVIDIA format

• M: The number of persons. The pre-trained model assumes a single object, but it can also support multiple people

The output of model inference is an array of N elements that gives the predicted action class for each sequence.

The labels used for training or evaluation are stored as a pickle file that consists of a list of two lists, including N elements each. The first list contains N strings of sample names. The second list contains the labeled action class ID of each sequence. The following is an example:

Copy
Copied!

[["xl6vmD0XBS0.json", "OkLnSMGCWSw.json", "IBopZFDKfYk.json", "HpoFylcrYT4.json", "mlAtn_zi0bY.json", ...], [235, 388, 326, 306, 105, ...]]


The graph to model skeletons is defined by two configuration parameters:

• graph_layout (string): Must be one the following candidates:

• nvidia consists of 34 joints. For more information, please refer to AR SDK Programming Guide.

• openpose consists of 18 joints. For more information, please refer to OpenPose.

• human3.6m consists of 17 joints. For more information, please refer to Human3.6M.

• ntu-rgb+d consists of 25 joints. For more information, please refer to NTU RGB+D.

• ntu_edge consists of 24 joints. For more information, please refer to NTU RGB+D.

• coco consists of 17 joints. For more information, please refer to COCO.

• graph_strategy (string): Must be one of the following candidates (for more information, refer to the “Partition Strategies” section in this paper):

• uniform: Uniform Labeling

• distance: Distance Partitioning

• spatial: Spatial Configuration

Note

All-in-one scripts are provided for processing Kinetics and self-annotated NVIDIA datasets. The preprocessed data and labels of the NVIDIA dataset can be accessed here.

## Creating an Experiment Spec File

The spec file for PoseClassificationNet includes model_config, train_config, and dataset_config parameters. Here is an example spec for training a 3D-pose-based model on the NVIDIA dataset. It contains 6 classes: “sitting_down”, “getting_up”, “sitting”, “standing”, “walking”, “jumping”:

Copy
Copied!

model_config:
model_type: ST-GCN
in_channels: 3
num_class: 6
dropout: 0.5
graph_layout: "nvidia"
graph_strategy: "spatial"
edge_importance_weighting: True
train_config:
optim:
lr: 0.1
momentum: 0.9
nesterov: True
weight_decay: 0.0001
lr_scheduler: "MultiStep"
lr_steps:
- 10
- 60
lr_decay: 0.1
epochs: 70
checkpoint_interval: 5
dataset_config:
train_data_path: "/path/to/train_data.npy"
train_label_path: "/path/to/train_label.pkl"
val_data_path: "/path/to/val_data.npy"
val_label_path: "/path/to/val_label.pkl"
label_map:
sitting_down: 0
getting_up: 1
sitting: 2
standing: 3
walking: 4
jumping: 5
batch_size: 16
workers: 5


 Parameter Data Type Default Description model_config dict config – The configuration for the model architecture train_config dict config – The configuration for the training process dataset_config dict config – The configuration for the dataset

### model_config

The model_config parameter provides options to change the PoseClassificationNet architecture.

Copy
Copied!

model_config:
model_type: ST-GCN
in_channels: 3
num_class: 6
dropout: 0.5
graph_layout: "nvidia"
graph_strategy: "spatial"
edge_importance_weighting: True


 Parameter Datatype Default Description Supported Values model_type string ST-GCN The type of model, which can only be ST-GCN for now. Newer architectures will be supported in the future. ST-GCN in_channels unsigned int 3 The number of input channels (dimension of body poses) >0 num_class unsigned int 6 The number of action classes >0 dropout float 0.5 The probability to drop hidden units 0.0 ~ 1.0 graph_layout string nvidia The layout of the graph for modeling skeletons. It can be nvidia, openpose, human3.6m, ntu-rgb+d, ntu_edge, or coco. nvidia/openpose/human3.6m/ntu-rgb+d/ntu_edge/coco graph_strategy string spatial The strategy of the graph for modeling skeletons. It can be uniform, distance, or spatial. uniform/distance/spatial edge_importance_weighting bool True Specifies whether to enable edge importance weighting True/False

### train_config

The train_config parameter defines the hyperparameters of the training process.

Copy
Copied!

train_config:
optim:
lr: 0.1
momentum: 0.9
nesterov: True
weight_decay: 0.0001
lr_scheduler: "MultiStep"
lr_steps:
- 10
- 60
lr_decay: 0.1
epochs: 70
checkpoint_interval: 5


 Parameter Datatype Default Description Supported Values optim dict config The configuration for the SGD optimizer, including the learning rate, learning scheduler, weight decay, etc. epochs unsigned int 70 The total number of epochs to run the experiment >0 checkpoint_interval unsigned int 5 The interval at which the checkpoints are saved >0 grad_clip float 0.0 The amount to clip the gradient by the L2 norm. A value of 0.0 specifies no clipping. >=0

#### optim

The optim parameter defines the config for the SGD optimizer in training, including the learning rate, learning scheduler, and weight decay.

Copy
Copied!

optim:
lr: 0.1
momentum: 0.9
nesterov: True
weight_decay: 0.0001
lr_scheduler: "MultiStep"
lr_steps:
- 10
- 60
lr_decay: 0.1


Parameter

Datatype

Default

Description

Supported Values

lr

float

0.1

The initial learning rate for the training

>0.0

momentum

float

0.9

The momentum for the SGD optimizer

>0.0

nesterov

bool

True

Specifies whether to enable Nesterov momentum.

True/False

weight_decay

float

1e-4

The weight decay coefficient

>0.0

lr_scheduler

string

MultiStep

The learning scheduler. Two schedulers are provided:
* MultiStep : Decrease the lr by lr_decay at setting steps.
* AutoReduce : Decrease the lr by lr_decay while lr_monitor doesn’t decline more than 0.1% of the previous value.

MultiStep/AutoReduce

lr_monitor

string

val_loss

The monitor value for the AutoReduce scheduler

val_loss/train_loss

patience

unsigned int

1

The number of epochs with no improvement, after which learning rate will be reduced

>0

min_lr

float

1e-4

The minimum learning rate in the training

>0.0

lr_steps

int list

[10, 60]

The steps to decrease the learning rate for the MultiStep scheduler

int list

lr_decay

float

0.1

The decreasing factor for the learning rate scheduler

>0.0

### dataset_config

The dataset_config parameter defines the dataset source, training batch size, and augmentation.

Copy
Copied!

dataset_config:
train_data_path: "/path/to/train/data.npy"
train_label_path: "/path/to/train_label.pkl"
val_data_path: "/path/to/val_data.npy"
val_label_path: "/path/to/val_label.pkl"
label_map:
sitting_down: 0
getting_up: 1
sitting: 2
standing: 3
walking: 4
jumping: 5
batch_size: 16
workers: 5


 Parameter Datatype Default Description Supported Values train_data_path string The path to the train data in a NumPy array train_label_path string The path to the train labels in a pickle file val_data_path string The path to the validation data in a NumPy array val_label_path string The path to the validation labels in a pickle file label_map dict A dict that maps the class names to indices random_choose bool False Specifies whether to randomly choose a portion of the input sequence. True/False random_move bool False Specifies whether to randomly move the input sequence. True/False window_size unsigned int -1 The length of the output sequence. -1 means the same as original length. batch_size unsigned int 64 The batch size for training and validation >0 workers unsigned int 1 The number of parallel workers processing data >0
Note

The input layout is NCTVM, where N is the batch size, C is the number of input channels, T is the sequence length, V is the number of keypoints, and M is the number of people.

## Training the Model

Use the following command to run PoseClassificationNet training:

Copy
Copied!

tao pose_classification train -e <experiment_spec_file>
-r <results_dir>
-k <key>
[gpu_ids=<gpu id list>]
[resume_training_checkpoint_path=<absolute path to \*.tlt checkpoint>]


### Required Arguments

• -e, --experiment_spec_file: The path to the experiment spec file.

• -r, --results_dir: The path to a folder where the experiment outputs should be written.

• -k, --key: The user-specific encoding key to save or load a .tlt model.

### Optional Arguments

• gpu_ids: The GPU indices list for training. If you set more than one GPU ID, multi-GPU training will be triggered automatically.

• resume_training_checkpoint_path: The path to a checkpoint to continue training.

Here’s an example of using the PoseClassificationNet training command:

Copy
Copied!



## Running Inference on the Model

Use the following command to run inference on PoseClassificationNet with the .tlt model.

Copy
Copied!

tao pose_classification inference -e <experiment_spec>
-k <key>
model=<inference model>
data=<path to inference data>
output_file=<path to output file>
[gpu_id=<gpu index>]


The output will be a text file, where each line corresponds to the predicted action class for an input sequence.

### Required Arguments

• -e, --experiment_spec: The experiment spec file to set up inference

• -k, --key: The encoding key for the .tlt model

• model: The .tlt model to perform inference with

• data: The path to the test data

• output_file: The path to the output text file

### Optional Argument

• gpu_id: The GPU index used to run the inference. You can specify the GPU index used to run inference when the machine has multiple GPUs installed. Note that inference can only run on a single GPU.

Here’s an example of using the PoseClassificationNet inference command:

Copy
Copied!



## Converting the Pose Data

Use the following command to convert the output JSON metadata from the deepstream-bodypose-3d app and generate spatio-temporal sequences of body poses for inference:

Copy
Copied!

tao pose_classification dataset_convert -k <key>
-e <experiment_spec>
data=<path to deepstream-bodypose-3d output data>
output_dir=<path to directory for output>
[pose_type=<pose type>]
[num_joints=<number of joints>]
[frame_width=<frame width>]
[frame_height=<frame height>]
[focal_length=<focal length>]
[sequence_length_max=<maximum sequence length>]
[sequence_length_min=<minimum sequence length>]
[sequence_length=<sequence length for sampling>]
[sequence_overlap=<sequence overlap for sampling>]


### Required Arguments

• -e, --experiment_spec: The experiment spec file to set up dataset conversion

• -k, --key: The encoding key for the .tlt model

• data: The output JSON data from the deepstream-bodypose-3d app

• output_dir: The directory for output

### Optional Arguments

• pose_type: The pose type can be chosen from 3dbp, 25dbp, 2dbp

• num_joints: The number of joint points in the graph layout

• frame_width: The width of frame images in pixels for normalization

• frame_height: The height of frame images in pixels for normalization

• focal_length: The focal length of the camera for normalization

• sequence_length_max: The maximum sequence length for defining array shape

• sequence_length_min: The minimum sequence length for filtering short sequences

• sequence_length: The general sequence length for sampling

• sequence_overlap: The overlap between sequences for sampling

Here’s an example of using the PoseClassificationNet dataset_convert command:

Copy
Copied!

tao pose_classification dataset_convert -e $DEFAULT_SPEC -k$YOUR_KEY data=$3D_BODYPOSE_JSON output_dir=$OUTPUT_DIR


The expected output would be a sampled array for each individual tracked ID saved under the output directory.

## Deploying the Model

You can deploy the trained deep learning and computer-vision models on edge devices, such as a Jetson Xavier, Jetson Nano, Tesla, or in the cloud with NVIDIA GPUs. The exported \*.etlt model can be used in the TAO Toolkit Triton Apps.

### Running PoseClassificationNet Inference on the Triton Sample

The TAO Toolkit Triton Apps provide an inference sample for Pose Classification. It consumes a TensorRT engine and supports running with either (1) a NumPy array of skeleton series or (2) output JSON metadata from the deepstream-bodypose-3d app.

To use this sample, you need to generate the TensorRT engine from an \*.etlt model using tao-converter.

#### Generating TensorRT Engine Using tao-converter

The tao-converter tool is provided with the TAO Toolkit to facilitate the deployment of TAO trained models on TensorRT and/or Deepstream. This section elaborates on how to generate a TensorRT engine using tao-converter.

For deployment platforms with an x86-based CPU and discrete GPUs, the tao-converter is distributed within the TAO docker. Therefore, we suggest using the docker to generate the engine. However, this requires that the user adhere to the same minor version of TensorRT as distributed with the docker. The TAO docker includes TensorRT version 8.0.

##### Instructions for x86

For an x86 platform with discrete GPUs, the default TAO package includes the tao-converter built for TensorRT 8.2.5.1 with CUDA 11.4 and CUDNN 8.2. However, for any other version of CUDA and TensorRT, please refer to the overview section for download. Once the tao-converter is downloaded, follow the instructions below to generate a TensorRT engine.

1. Unzip the zip file on the target machine.

2. Install the OpenSSL package using the command:

Copy
Copied!

sudo apt-get install libssl-dev


3. Export the following environment variables:

Copy
Copied!

$export TRT_LIB_PATH=”/usr/lib/x86_64-linux-gnu”$ export TRT_INC_PATH=”/usr/include/x86_64-linux-gnu”


1. Run the tao-converter using the sample command below and generate the engine.

2. Instructions to build TensorRT OSS on Jetson can be found in the TensorRT OSS on x86 section above or in this GitHub repo.

Note

Make sure to follow the output node names as mentioned in the Exporting the Model section of the respective model.

##### Instructions for Jetson

For the Jetson platform, the tao-converter is available to download in the NVIDIA developer zone. You may choose the version you wish to download as listed in the overview section. Once the tao-converter is downloaded, please follow the instructions below to generate a TensorRT engine.

1. Unzip the zip file on the target machine.

2. Install the OpenSSL package using the command:

Copy
Copied!

sudo apt-get install libssl-dev


3. Export the following environment variables:

Copy
Copied!

$export TRT_LIB_PATH=”/usr/lib/aarch64-linux-gnu”$ export TRT_INC_PATH=”/usr/include/aarch64-linux-gnu”


1. For Jetson devices, TensorRT comes pre-installed with Jetpack. If you are using older JetPack, upgrade to JetPack-5.0DP.

2. Instructions to build TensorRT OSS on Jetson can be found in the TensorRT OSS on Jetson (ARM64) section above or in this GitHub repo.

3. Run the tao-converter using the sample command below and generate the engine.

Note

Make sure to follow the output node names as mentioned in Exporting the Model section of the respective model.

##### Using the tao-converter

Here is a sample command to generate the PoseClassificationNet engine through tao-converter:

Copy
Copied!

#convert ST-GCN model with 3D poses, input sequence length of 300, and 34 keypoints:
tao-converter <etlt_model> \
-k <key_to_etlt_model> \
-d 3,300,34,1 \
-p input,1x3x300x34x1,4x3x300x34x1,16x3x300x34x1 \
-o fc_pred \
-t fp16 \
-m 16 \
-e <path_to_generated_trt_engine>


This command will generate an optimized TensorRT engine.

#### Running the Triton Inference Sample

You can generate the TensorRT engine when starting the Triton server using the following command:

Copy
Copied!

bash scripts/start_server.sh


When the server is running, you can get results from a NumPy array of test data with the client using the command mentioned below:

Copy
Copied!

python tao_client.py <path_to_test_data> \
-m pose_classification_tao \
-x 1 \
-b 1 \
--mode Pose_classification \
-i https \
-u localhost:8000 \
--async \
--output_path <path_to_output_directory>


Note

The server will perform inference on the input test data. The results are saved as a text file where each line is formatted as [sequence_index], [rank1_pred_score]([rank1_class_index])=[rank1_class_name], [rank2_pred_score]([rank2_class_index])=[rank2_class_name], ..., [rankN_pred_score]([rankN_class_index])=[rankN_class_name]. The expected output for the NVIDIA test data would be as follows:

Copy
Copied!

0, 27.6388(2)=sitting, 12.0806(3)=standing, 7.0409(1)=getting_up, -3.4164(0)=sitting_down, -16.4449(4)=walking, -26.9046(5)=jumping
1, 21.5809(2)=sitting, 8.4994(3)=standing, 5.1917(1)=getting_up, -2.3813(0)=sitting_down, -12.4322(4)=walking, -20.4436(5)=jumping
2, 5.6206(0)=sitting_down, 4.7264(4)=walking, -1.0996(5)=jumping, -2.3501(1)=getting_up, -3.2933(3)=standing, -3.5337(2)=sitting
....


You can also get inference results from the JSON output of the deepstream-bodypose-3d app using the following command:

Copy
Copied!

python tao_client.py <path_to_json_file> \
--dataset_convert_config ../dataset_convert_specs/dataset_convert_config_pose_classification.yaml \
-m pose_classification_tao \
-x 1 \
-b 1 \
--mode Pose_classification \
-i https \
-u localhost:8000 \
--async \
--output_path <path_to_output_directory>


Note
The server will perform inference on the input JSON file. The results are also saved as a JSON

file, which follows the same format as the input and adds the predicted "action" to each object at each frame. A sample of the JSON output would be as follows:

The skeleton sequence of each object is broken into segments by a dataset converter (refer to the figure below). The sequence_length and sequence_overlap are configurable in dataset_convert_config_pose_classification.yaml. The output labels are assigned to frames after a certain period of time.

#### End-to-End Inference Using Triton

A sample for end-to-end inference from video is also provided in the TAO Toolkit Triton Apps. The sample runs deepstream-bodypose-3d to generate metadata of bounding boxes, tracked IDs, and 2D/3D poses that are saved in JSON format. The client implicitly converts the metadata into arrays of skeleton sequences and sends them to the Triton server. The predicted action for each sequence is returned and appended to the JSON metadata at corresponding frames. A video with overlaid metadata is also generated for visualization.

You can start the Triton server using the following command (only the Pose Classification model will be downloaded and converted into a TensorRT engine):

Copy
Copied!

bash scripts/pose_cls_e2e_inference/start_server.sh


Once the Triton server has started, open up another terminal and run the following command to begin body pose estimation using DeepStream and run Pose Classification on the DeepStream output using the Triton server instance that you previously spun up:

Copy
Copied!

bash scripts/pose_cls_e2e_inference/start_client.sh