Nodeflux: Automating computer vision analysis for the smart cities of tomorrow
Indonesia · Oct 05, 2018· By Putu Agung Wija Putera
Indonesia's first facial recognition tech company aims to become a central player in Jakarta's smart city push
Fans of spy thrillers like the Bourne movies or Person of Interest will be familiar with the concept of advanced AI that analyzes CCTV footage to recognize faces, track movements and perhaps detect criminal activity in real time. Today, such computer vision technology is being actively developed and applied by police and security agencies around the world.
Computer vision technology can be expensive to apply, not to mention controversial. Indonesian startup Nodeflux, however, is confident it can bring this advanced technology to the local market. Nodeflux offers software and infrastructure that enables companies and authorities to analyze real-time footage using machine learning capabilities.
Aside from facial recognition, analysis can be extended to traffic management and retail optimization. The potential of Nodeflux's platform has not gone unnoticed, and the company has received funding from smart city provider Qlue and prominent VC fund East Ventures.
Nodeflux started out by offering what it called a 'distributed computing' platform. Simply put, the platform allows users to process and coordinate data from multiple, separate devices such as cameras, microphones or sensors. According to the company, Nodeflux’s platform can be used to process text, images, audio and video data.
The Nodeflux platform operates by processing data in separate nodes. Each node represents one process in the data analytics workflow. Interestingly, different nodes in the workflow can be programmed using different coding languages, depending on what is most efficient for the required purpose. The nodes can also be connected so that output from one node is later processed in another.
Initially, Nodeflux sought to incorporate AI concepts such as deep learning to process data inputs. The company eventually focused on processing a certain type of data: videos. But it did not abandon its original distributed computing concept; rather it chose to limit its scope to processing video data from devices like CCTV.
Smart city applications
Obvious security applications include facial recognition and vehicle license plate recording. Nodeflux also offers other applications for smart city management and even retail. For example, smart city managers can use image recognition to assess traffic flows or detect parking violations. Retailers can analyze customer footfall in stores to determine which displays or products are attracting the most attention.
Nodeflux has demonstrated its capabilities through several partnerships. During the 2018 Eid al-Fitr holiday rush, for example, Nodeflux partnered with state-owned toll road operator Jasa Marga. Nodeflux created a system for counting vehicles entering and exiting rest stops, which was then used to calculate and display the number of empty parking spots in real time. This enabled Jasa Marga to improve management of rest stop traffic.
Nodeflux is also a partner of Jakarta’s smart city initiative, which also involves Qlue. One of the earliest applications of the Nodeflux platform is detecting crowd congestion levels in the city’s Transjakarta bus shelters. The company also names Bandung’s smart city initiative, the national police and the Indonesian intelligence agency as clients.
Perhaps Nodeflux’s most significant partnership to date is with GPU (graphics processing unit) producer NVIDIA. Nodeflux claims to be the first Indonesian startup to join the NVIDIA Inception Program, an accelerator for AI and deep learning startups. By using NVIDIA’s GPU to process deep learning tasks instead of just the CPU (central processing unit), Nodeflux can offer faster video analysis than previously.
Nodeflux is far from being the only company in the world offering computer vision solutions, but its advantage perhaps lies in how quickly it can secure local partnerships. Whether or not the company has the potential to expand regionally remains to be seen, but Indonesia remains a large enough market for them to explore and profit from and has many cities and companies that could benefit from CCTV footage analysis.
Nodeflux must quickly demonstrate the benefits of automating image and video analysis in multiple contexts before competing startups pick up steam or regional players start entering the Indonesian market. By securing brand-name recognition and overall market share in Indonesia, Nodeflux could stand proudly as Indonesia’s computer vision pioneer.
Edited by Sophie Douez