Since the traditional way of textile inspection, which is putting all the responsibility on human’s eyes to maintain the yield rate of textile, is not suitable for a variety of technical textile, Taiwan Textile Research Institute (TTRI) has developed an Automated AI Textile Inspection Solution, based on NVIDIA Jetson-Based Edge AI Computing Platform and the big data that TTRI has collected for years.
Consists of NVIDIA Jetson-based platform with Millitronic’s 5G WiGig module, TTRI’s textile inspection algorithm and defect management software, this automated textile inspection solution provides comprehensive I/O function on-board and supports driver-ready extension I/O modules, which is suitable for the traditional weaving producing line and provides intranet environment through 60GHz WiGig signal, fulfilling the high-speed communication and real-time management between equipment and control center. In the same time, TTRI conquered the ordeals, the variety of textile and defect, through AI deep learning, which can keep learning and updating while doing AI inference.
In addition, NVIDIA Jetson-based platform provides powerful GPU parallel computing ability, which enable to do real-time inferring with the defect image and sent the results to the back-end software for management. As the result, TTRI is allowed to utilize the data for textile reducing process optimization and new textile development, and push the textile industrial into a highly precision and technology-intensive AI industrial.
|Flexible Deployment||Defect Software||Mutiple Cameras|
|User-friendly and easy to install||Training AI deep learning model||Support multiple cameras|
|Aetina Jetson Carrier Board : ACE-N622|
|• Support driver-ready extension peripheral modules for I/O expansion
• On-board 1x HDMI/2x CANBus/4x GPIO/1x RS232/1x I2C/2x Full-Mini Cards
• Support extended temperature -20°C to 70°C
|Hardware Peripherals||Software & Algorithm|
|Module : NVIDIA Jetson TX2||Algorithm : TTRI Textile Algorithm|
|VID Module : Industrial Camera||Software : Defect Software|
|Millitronic WiGig Module||Edge Database|
|Connectivity through ACE-N622|
|Software & Algorithm|
TTRI Textile Algorithm• Defects : broken yarn/shrinked weft/broken ends/felter/start-up mark/end out/missing pick/stain(dirt)
• Defect inference accuracy is increased up to 90%
Millitronic WiGig Module• Advanced 60GHz gigabit distribution network products
• Versatile and fast deployment enabled by edge and cloud tools