Mobiliya ADAS

ADAS in automotive industry has been around for a while now and is gaining popularity as the industry is moving towards connected car and autonomous driving realization. Mobiliya, with its strong foundation, in Deep Learning (DL) is building ADAS solutions which span beyond simple object detection to more contextualized content interpretation. This essentially means that instead of passive safety measures that are restricted to just minimizing accidents, we are now moving towards more enhanced ADAS that avoid accidents completely. This includes blind spot detection that can alert a driver as he tries to move into an occupied lane, lane departure warning, alerting the driver if the car is drifting outside its lane, pedestrian detection and distinguishing a police car from an ambulance. Mobiliya aims to partner with OEMs, Tier 1 suppliers, automotive research groups and start-ups to develop and deploy breakthrough autonomous vehicles using Deep Learning.


Traffic Sign

Traffic Sign Detection Using DL for ADAS

One of the biggest reasons for road accidents is because drivers often tend to overlook traffic signs and warnings. Thus, to make vehicles more sensitive to traffic signs and warnings, Mobiliya offers a 5-step process to train a Deep Learning network for detecting traffic signs:

  • Deploy a suitable DL network. For example YOLO, DetectNet
  • Identify a suitable traffic signs dataset from publicly available datasets like the German Traffic Sign Detection Benchmark (GTSDB), Laboratory for Intelligent & Safe Automobiles (LISA)n
  • Train the DL network using any of these datasets
  • Do extensive testing of the network using test dataset and images in the wild
  • Do continuous tuning and retraining of the network based on the results
Mobiliya ADAS Offering 1
Mobiliya ADAS Offering 2
Smart Parking

Smart Parking

A research conducted by UCLA suggests that about 30% of downtown traffic merely consists of drivers looking a place to park. The study suggests that the average time a driver spent for parking was 3.3 min while the average distance covered was half-mile. This also meant that annually, a search for parking around just the LA campus would add up to 950,000 miles of travel with a fuel consumption of 47,000 gallons and 730 tons of greenhouse emissions.

All this can be reduced tremendously with smart parking, with drivers tending to spend only half as much time looking for a parking space.

Mobiliya’s Smart Parking solution offers identification of available parking slot counting in real time. It also provides accurate information on vehicle in/out time and can be easily integrated with ticket vending machine and payment management.

Mobiliya’s Smart Parking solution is based on:

  • Use of video cameras to detect parking spot occupancy
  • Obtaining car image datasets
  • Training and validating pre-trained DetectNet model available with NVIDIA DIGITS and KITTI dataset
  • Performing inference and finding accuracy
  • Deploying the network on Jetson