Machine Learning is a current application of AI based around the idea that we should just be able to give machines access to data and let them learn for themselves.
New technologies and capabilities have emerged to enable the development of intelligent decision-making solutions from data provided to compute systems. For multiple applications and industries, there is a close working relationship between AI, Machine Learning, Neural Networks and Deep Learning. Development platforms are available for Machine Learning / Deep Learning developments including TensorFlow and Caffe.
Deep Learning is one of the most effective methods for recognizing patterns and developing insights from unstructured data such as images, sounds, video and text, and is a key branch of Machine Learning & Artificial Intelligence development. Applications include Image Recognition, Behaviour Data Analysis, Video Surveillance and Automation & Robotics.
Deep Neural Networks (DNN) are ‘trained’ with large datasets to ‘learn’ to classify or detect objects. In Computer Vision Applications Convolutional Neural Networks (CNN) are used to maximise the visual recognition.
The emergence of AI/ Machine Learning/Neural Networks/Deep Learning solutions has also challenged the hardware side with fast and efficient SoC/ASIC/GPU/FPGA devices and solutions required.
Chipright has delivered design services within the ML technological domain by utilizing a proven pool of highly skilled senior level engineering resources. Chipright provides engineers and teams that are expert in their fields on-site or remotely. We have the capacity to supply the market with the right skill at the right price in the right location at the right time.
New clients often ask us about projects we have worked on and implemented over time in the ML market space. They also ask us about the resource pool we use on these projects. We respect the curiosity but also respect our clients NDA’s. Thus, whilst we are restricted from conveying specific information about the R&D technological projects we continually work on, we can provide a brief snapshot of some of the work without disclosing our customers detailed project information here.
Case Study 1 – SOC Verification of Convolutional Neural Network Engine with ARC processors
- Design and develop a C-based verification environment
- MLV test cases in C for TNN ASIP processors.(Dual Processor core)
- Enhanced thread management
- Memory management with DMA-AXI transfers
- Writing C test cases for SOC environment from scratch for ARC HS cluster (4 ARC processor +APEX for user interface)
- Randomization to randomize the image pixel values, Convolution co-efficient for CNN engine
- Checking the image processing algorithms ,Techniques involved in image processing to handle it for Convolutional neural network
Case Study 2 Verification of a mixed-signal SoC design with a Cortex-M4, a GPU / Neural Network Engine and multiple custom blocks
- Implemented a pragmatic verification environment that allowed usage of available UVM UVCs
- Support for legacy embedded software to enable communication and co-ordination between verification environment and embedded software
- Extracted functional and connectivity focused verification requirements from available sources
- Definition and implemented cluster and top level directed tests
- Performed tracking of verification progress
- Trained engineers at the client’s end to leverage best in class UVM base class functionality – to maintain the infrastructure longer term for the client