Sitara Machine Learning
What is Sitara Machine Learning?
With an increased push toward industry 4.0, industrial equipment are becoming increasingly connected. Their operational data captured, clustered and analyzed for trends to help increase manufacturing efficiency. This is accomplished my enabling machine learning at the edge.
All Sitara devices enable deep learning at the edge by running machine learning inference. Inference is a trained neural network that has been deployed into a processor to infer on new input data, based on its training.
Some examples include: inspecting parts on an industrial production line, counting and tracking people within a room, predicting the remaining useable life of a piece of equipment, or determining whether a currency is counterfeit.
Advantages of using Sitara Processor for ML Applications in Smart Factories
- Integrated SoC with key industrial peripherals to provide the processing and connectivity needed by smart factories products
- Industrial protocol support.
- Real-time control and processing with PRU and C66x DSP.
- Accelerated Deep Learning for high performance and low power.
- CSI-2 and parallel interface for camera connection.
- Tons of connectivity: PCIe, USB3, SATA and more.
- Enhanced Reliability with extended temperature & high voltage I/O, Low Failure in Time (FIT) and high POH (Power On Hours) with device longevity.
- Lower system cost
- Just enough performance to enable running inference at the edge for many machine vision applications.
- Single chip solution for field, controller or operator level devices along with PdM at the edge.
- Lower system power, < 5W.
- Low Latency.
- Scalable family with pin-compatible single and dual core devices with single Processor SDK for all Sitara devices.
What are some example applications where Sitara Deep learning can be used?
- Industrial robots for automated sorting.
- Vision computer & optical Inspection.
- Automated Guided Vehicle.
- Smart HMI.
- Optimize equipment settings.
- Track/identify/count people and objects.
- Extracting useful information gathered from the data aggregator.
- ATM currency counter.
- Predictive Maintenance (PdM) (Identify anomalies, machine wear and expected lifetime of equipment).
- Classify and recognize specific sounds/audio patterns.
Getting started with the evaluation of Sitara Machine learning
The reference design Deep Learning Inference for Embedded Applications Reference Design demonstrates TI deep learning (TIDL) on a Sitara AM57x. Below are the hardware and software needs to get started.
* Of the entire Sitara portfolio, AM5729 is a highest performing processor for machine learning.
+ Machine learning functions performed on the C66x DSP cores.
TI Deep Learning (TIDL) software
TIDL software leverages a highly optimized neural network implementation on TI’s Sitara AM57x processors, making use of hardware acceleration on the device. TIDL is a set of open-source Linux software packages and tools that enables offloading of deep learning inference to either Embedded Vision Engine (EVE) subsystems, C66x DSPs, or both. TIDL software is available as part of TI’s free AM57x Linux Processor SDK.
Other Helpful Resources:
Bringing machine learning to embedded systems (white paper)