Search results
View Joohoon Lee’s profile on LinkedIn, a professional community of 1 billion members. I lead the vision, strategy, and execution of cutting-edge solutions for various domains,…
- 500+
- Director of Product @ NVIDIA | AI/ML Software
- 1.9K
Lee Je-hoon ( Korean : 이제훈; born July 4, 1984) is a South Korean actor. He started his career in indie films.
- The Cityscapes Dataset and Fully Convolutional Network
- Int8 Inference and Calibration
- Creating A Calibration Cache Using The Python API
- Optimizing The Int8 Model on Drive Px
- Get Your Hands on TensorRT 3
The Cityscapes Dataset[Cordts et al. 2016] is designed for semantic segmentation of urban autonomous driving scenarios. Figure 2 shows sample images from the dataset. The dataset has a total of 30 different classes, grouped into 8 different categories. For the purpose of evaluating performance, I use 19 classes and 7 categories, as Figure 3 shows. ...
The Pascal dGPU in DRIVE PX AutoChauffeur is capable of executing 8-bit integer 4-element vector dot product (DP4A, see Figure 6) instructions to accelerate deep neural networkinference. While this new instruction provides faster computation, there is a significant challenge in representing weights and activations of deep neural networks in this re...
With the introduction of the TensorRT Python API, it is now possible to implement the INT8 calibrator class purely in Python. This example shows how to process image data and feed the calibrator. It should be simple to modify this example to support different types of data and networks in Python. The PythonEntropyCalibrator class is a Python implem...
TensorRT builder implements a profiling-based optimization called kernel autotuning. This process requires the network to be optimized on the target device. We can use the calibration cache file generated from the host in this on-target optimization phase to generate an INT8 model without requiring the calibration dataset. You need to write a calib...
NVIDIA TensorRT is a high-performance deep learning inference accelerator that delivers low latency, high-throughput inference for deep neural networks. TensorRT 3 is available now for x86 systems. TensorRT 3 will be available for DRIVE PX 2 by mid-January 2018. If you would like to get it sooner, please contact your NVIDIA solution architect or re...
Dec 12, 2017 · In a new NVIDIA Developer Blog post by Joohoon Lee, the lead of the Automotive Deep Learning Solutions Architect team at NVIDIA, shows you how to use the TensorRT 3 Python API on the host to cache calibration results for a semantic segmentation network for deployment using INT8 precision.
- Brad Nemire
Joohoon Lee’s Post. Manager at LS ELECTRIC. 4mo. The arrangement is expected to strengthen LS MnM’s secondary battery materials business. Formerly the CEO of E1 Corp, a liquefied...
Dec 11, 2017 · Joohoon Lee leads the Automotive Deep Learning Solutions Architect team at NVIDIA. He focuses on turning deep learning research into real world autonomous driving software for production deployment.
Joohoon Lee is a principal product manager for AI inference software at NVIDIA. Previously, he led the automotive deep learning software solutions team focusing on the production deployment of neural networks in DRIVE AGX platform using TensorRT.
