Scientist: Video/Multimodal Data Analytics
Scientist: Video/Multimodal Data Analytics (Job Number: P17F05)
Mountain View, CA

This position offers the opportunity to conduct innovative research on a broad set of problems related to multi-modal temporal segmentation.

Key Responsibilities:

 

  • Propose, create, and implement supervised and unsupervised data segmentation/clustering algorithms from multimodal and multisensory data streams obtained from traffic scenes.
  • Develop and evaluate metrics to verify reliability of the proposed algorithms.
  • Participate in ideation, creation, and evaluation of related technologies in various domains other than traffic scenes, including temporal segmentation of human activities.
  • Contribute to a portfolio of patents, academic publications, and prototypes to demonstrate research value.
  • Participate in data collection, sensor calibration, and data processing.
  • Participate in software development and implementation on various experimental platforms.

Qualifications:

 

  • PhD in computer science, electrical engineering, or related field.
  • Research experience in computer vision, machine learning, and multi-modal signal processing.
  • Strong familiarity with machine learning techniques pertaining to sequential data processing.
  • Preferred hands on experience in handling multi-modal sensor data.
  • Preferred experience in open-source Deep Learning frameworks such as TensorFlow or Caffe.
  • Highly proficient in software engineering using C++ and Python.
  • Strong written and oral communication skills including development and delivery of presentations, proposals, and technical documents.
  • Strong publication record in one or more of the following areas: computer vision, machine learning, or computer vision.

Projects

We compute various musculoskeletal indicators of human performance when the driver is operating a vehicle under normal and emergency maneuvering.
The goal of the project is to achieve robust lane-level localization for cars using low cost mass producible sensors.
We have developed online algorithms to transfer motion from a human demonstrator to Honda's humanoid robot, ASIMO.
The development of the next generation green and safe batteries with high energy density is highly desirable for meeting the rapidly growing needs of electrical vehicles.
Rare earth Manganese Oxide (ReMO) as a cathode material has potential capacity higher than that of commercial lithium manganese oxide.
Metal-air batteries are solid state batteries using metal oxidation at the anode and oxygen reduction at the cathode to induce a current flow.
This project is focused on CNT and graphene enforced electrode materials for secondary battery and supercapacitor applications.
This research is focused on scale-up technology for continuous synthesis of SWNTs by CVD method and the exploration of their performance in actual electrochemical devices
The core of this project is for atomistic level understanding of environmental impact on conductivities of SWNTs and 2-D materials in order to reveal their ultimate sensitivities.
Synthesis and studies of growth mechanism, self assembly and properties of low dimensional nanomaterials for alternative energy technologies are at the core of our research.
The goal is to develop driving aids that enhance the driver's situational awareness and give drivers a sense of confidence and trust in the vehicles they are operating.
In Knowledge Discovery, we Integrate knowledge from multiple sources such as Wikipedia, Yahoo Question/Answers, Open Directory Project and OpenMind.
Probabilistic model to track dialog state and provide information to driver viaspoken dialog
This project presents a control theoretic approach for human pose estimation from a set of key feature points detected using depth image streams obtained from a time of flight imaging device.
We are developing a real-time system that detects and tracks traffic participants.
We aim to develop a robust pedestrian detection algorithm that can handle partial occlussions.
We are taking advantage of our autonomous driving platform to create a comprehensive repository of annotated sensor data that provide computer vision benchmarks and training data to support advanced driving assist and autonomous driving applications.
Backing-up of articulated vehicles poses a difficult challenge even for experienced drivers. While long wheelbase dual-axle trailers provide a benefit of increased capacity over their single-axle counterparts, backing-up of such systems is especially difficult. We devise a control strategy for such systems, allowing backing-up maneuvers to be intuitive to drivers without experience with trailers. Using hitch angle feedback, we show these concepts can be used to stabilize the trailer in back-up motion in the presence of arbitrary driver inputs.
Utilizing the latest wearable sensing technologies and patented motion prediction algorithms, the goal is to predict human movement and perform biomechanical computations based on those predictions.