NVIDIA Transfer Learning Toolkit

Create accurate and efficient AI models for Intelligent Video Analytics and Computer Vision without expertise in AI frameworks. Develop like a pro with zero coding.


Get Started


Transfer learning extracts learned features from an existing neural network to a new one. Transfer learning is often used when creating a large training dataset is not feasible. Transfer Learning Toolkit (TLT) is a python based AI toolkit for taking purpose-built pre-trained AI models and customizing them with your own data. Developers, researchers and software partners building intelligent vision AI apps and services, can bring their own data to fine-tune pre-trained models instead of going through the hassle of training from scratch.



Easier & Faster Training

Add state of the art AI to your application with zero coding. No AI frameworks expertise needed

Highly Accurate AI

Remove barriers and unlock higher network accuracy by using purpose-built pre-trained models

Greater Throughput

Reduce deployment costs significantly and perform high throughput inference with DeepStream SDK and TLT



Transfer Learning Toolkit adapts popular network architectures and backbones to your data, allowing you to train, fine tune, prune and export highly optimized and accurate AI models for edge deployment. TLT supports commonly used DNNs for image classification and object detection .


Image Classification
Object Detection
Instance Segmentation
DetectNet_V2
FasterRCNN
SSD
YOLOV3
RetinaNet
DSSD
MaskRCNN
ResNet
10/18/34/50/101

VGG16/19

GoogLeNet

MobileNet V1/V2

DarkNet 19/53

SqueezeNet

+ - MaskRCNN segmentation network, coming soon in TLT 2.0 GA


Purpose-built Pre-trained Models

The pre-trained models accelerate the AI training process and reduce costs associated with large scale data collection, labeling, and training models from scratch. The Purpose-built pre-trained models are highly accurate and can be used for common object detection and image classification use-cases such as counting people, detecting vehicles, optimizing traffic, parking management and more.


Model
Network Architecture
Number of Classes
Accuracy
Use-Case
DashCamNet
DetectNet_v2-ResNet18
4
80%
Identify Objects from a moving object like car or robot
FaceDetect-IR
DetectNet_v2-ResNet18
1
96%
Detects face in a dark environment close to camera
PeopleNet
DetectNet_v2-ResNet34
3
84%
People counting, heatmap generation, social distancing
TrafficCamNet
DetectNet_v2-ResNet18
4
83.5%
Detect and track cars
VehicleMakeNet
ResNet18
20
91%
Classifying car models
VehicleTypeNet
ResNet18
6
96%
Classifying cars in a parking management or toll booth

Purpose-built models from TLT when deployed with DeepStream SDK, enables you to unlock greater end-to-end throughput.



Jetson Nano
Jetson Xavier NX
Jetson AGX Xavier
T4
Model Architecture
Inference Resolution
Precision
Model Accuracy
GPU (FPS)
GPU (FPS)
DLA1 (FPS)
DLA2 (FPS)
GPU (FPS)
DLA1 (FPS)
DLA2 (FPS)
GPU (FPS)
PeopleNet-ResNet34
960x544
FP16
84%
10
60
30
30
120
60
60
460
TrafficCamNet-ResNet18
960x544
INT8
83.5%
19*
180
90
90
420
120
120
1300
DashCamNet-ResNet18
960x544
INT8
80%
18*
180
90
90
390
120
120
1280
FaceDetect-IR-ResNet18
384x240
INT8
96%
95*
1080
570
570
1950
780
780
2160
Tabulated data is in FPS with 1080p input
* FP16 inference on Jetson Nano
Running on the DLAs for AGX Xavier and NX frees up GPU for other tasks


PeopleNet


Identify foot traffic to operationalize retail stores, malls and mass transit locations. PeopleNet is a 3-class object detection network trained on 960x544 RGB images to detect person, bag and face.

TrafficCamNet & VehicleMakeNet


Understand traffic flow around intersections and optimize traffic during congestion with TrafficNet, a 4-class object detection network trained on 960x544 RGB images to detect cars, two wheelers, persons, and road signs.


There are 25+ TLT pre-trained models for objection detection and classification available on NVIDIA NGC

See All Models


Why Use Transfer Learning Toolkit?



Simplified Vision AI

  • Faster time to market with zero coding and production-ready DNNs
  • Simple command line interface (CLI) abstracts AI framework complexity
  • Readily deployable purpose-built pre-trained models
  • Speed up model training with multiple-GPUs





Create End-to-End AI Systems

  • Build end-to-end services and solutions for transforming pixels and sensor data to actionable insights using TLT, DeepStream SDK and TensorRT
  • TLT pruning improves channel density for high throughput inference
  • Use TLT on NVIDIA DGX, cloud GPU instances or your local workstations


General FAQ

Yes, TLT models are free for commercial use. For specific licensing terms, refer to model EULA.

TLT uses TensorFlow and Keras framework completely abstracted away from the user. Users operate TLT through documented spec files and do not have to learn about DL framework.

Pull the TLT container from NGC. The container comes pre-packaged with Jupyter notebooks and sample spec files for various network architectures. Additional technical resources can be found here.

No third party pre-trained models are supported by TLT. Only NVIDIA pre-trained models from NGC are currently supported.

Training with TLT is only on x86 with NVIDIA GPU such as a V100. Models trained with TLT can be deployed on any NVIDIA platform including Jetson.

To deploy trained models on DeepStream, refer to Deploying to DeepStream chapter of TLT Getting started guide.

The six purpose-built models (PeopleNet, TrafficCamNet, DashCamNet, FaceDetectIR, VehicleTypeNet and VehicleCamNet) can be used as is out of box and can also be re-trained with your dataset. The architecture specific models for detection and classification are required to be re-trained with TLT.

Segmentation is currently not supported and it will be available in the coming GA release. See the full model support matrix above.

QAT (Quantization Aware Training) to improve INT8 accuracy and support for AMP to speed up AI training with tensor cores enabled for NVIDIA GPUs will be coming in Q3, 2020 with TLT 2.0 general availability.

Latest Product News




Developer Tutorial

Get started with training using custom pre-trained models and TLT, export models for deployment with DeepStream


Learn More


Developer Webinar

Learn about the end-to-end AI workflow using Transfer Learning Toolkit and GPU-optimized pre-trained models.


Learn More


NVIDIA GTC

BMW research group showcases use of NVIDIA ISAAC SDK and TLT for building smart transport robots.


Learn More


Community Projects

Learn something new or build your own project. Explore innovative projects built by our developer community.


Submit A Project


Simplify and speed-up AI training with Transfer Learning Toolkit


Get Started