The global AI in textile market is valued at around $2.4 billion in 2023 and is projected to reach nearly $21.4 billion by 2033, growing at a CAGR of over 24%. Yet on the factory floor, the conversation is far more nuanced than market projections suggest.
In the fourth episode of ReCIRCLE by TexSPACE, industry experts with hands-on experience in knitting production, AI fabric inspection, and machine health systems come together to discuss one focused question: how ready is AI for the knitting industry, and where does it still fall short?
This episode features Eng. Md. Aminul Islam, Assistant General Manager (R&D Textile) at Metro Dyeing and Knitting Ltd.; Jaivardhan T, Director of Sales at Count AI; and Labeeb Hossain, Director of Sales and Marketing at Texcorp BD.
AI can detect defects on a large scale, but it is not “magic”
Director sales at count AI
Jaivardhan explains that AI-based fabric inspection systems have made significant progress in the last four years, especially in the field of online knitting inspection. However, he emphasizes that the effectiveness of AI depends on how visible the defect is inside the machine. If a defect is not visible due to fabric tension, low visual contrast or complex structure, then even AI will not be able to detect it.
He said, “AI is a tool. AI cannot do magic, but AI can enhance what the human can do on a much larger scale.”
He points out that very fine, low-contrast defects hidden within the fabric structure are still challenging for AI to detect. However, he believes that even if AI can currently provide only partial solutions, factories should still adopt the technology. If AI can detect 80 out of 100 defects, it represents a major improvement over fully manual inspection and can significantly reduce labor dependency and human error. As AI systems continue learning from data, they will gradually become capable of detecting the remaining 20 defects as well
He highlighted it as a broader production intelligence tool. By analyzing machine performance, yarn quality, needle brand, operator skill, and order type, AI can help determine the best production setup for a specific customer order. However, which tasks to prioritize and what strategic decisions to make are still dependent on human judgment. AI cannot make these decisions.
He said, “AI is always going to be in a developing phase. There is always scope for improvement. It’s never going to be 100%.”
Practical factory experience highlights both AI’s promise and its current limitations
Aminul Islam brings a perspective that doesn’t often get heard in product demos. His factory has trialed AI quality systems, and the experience has been a mixed one.
The most pressing issue isn’t missed defects, it’s false ones. In some variegated rib or needle-drop designs, what looks like a fault to the AI camera is actually part of the intended pattern. Every time the system flags it and stops the machine, production takes a hit. With margins already thin, that adds up.
“The harsh truth is that when a company owner invests in this technology, and the AI system hasn’t matured enough for production output, the management gets disappointed,” he says. “Whoever is bringing this technology into our country should first run it as a prototype or a small project. Understand the limitations. Then invite feedback. Only then make the decision.”
He also points to specific defects the system has failed to catch – line marks and sinker marks among them and notes that even shade variation within the same lot of yarn has triggered false alarms.
His suggestion is a collaborative one: “If the developers and the knitting experts work together, they can move this forward. Otherwise it will be very difficult to convince factory owners and management.”
One recurring challenge in the discussion is AI flagging design features as defects – the most-cited example being pointelle fabric, where small holes are part of the intended structure.
The solution, Labeeb Hossain explains, is giving the system a reference point. “You feed the AI the actual pointelle fabric. It sees this and realizes – this is what normal looks like. So anything different from this pattern is a problem. But if holes appear in specific patterns all over the fabric, that’s okay, that’s the design. It needs to know what normal is before it can find the abnormal.“
Predictive maintenance can shift factories from reactive to proactive operations
Most factories today still operate on reactive maintenance. Something breaks, then it gets fixed. Labeeb Hossain’s work is built around changing that.
“What we are trying to do is fix my machine before there’s a problem. These two are very big distinctions,“ he says.
The system his company works with collects real-time data from needles, sinkers, cylinders, motors, feeders, and vibration sensors across the circular knitting machine. A predictive AI model then analyzes this data against historical maintenance records to flag machines running outside normal parameters.
He gives a concrete example: “The machine is supposed to run at 36°C. For some reason it’s running at 40°C. It’s not possible for a human to find that 4° change. But the system can. And when it sees that, it can tell you either your lubrication is defective or your motor has a lot of resistance.“
For an industry where every minute of downtime means lost money, the goal is simple: keep the machine running.
Beyond defect detection, Jaivardhan sees AI playing a wider role in production planning – one that’s less visible but potentially more valuable.
He describes a scenario where a factory manager with 100 knitting machines, ten yarn suppliers, multiple needle brands, and twenty operators receives a priority order from a brand. With full operational data, AI can recommend the optimal combination of machine, operator, needle, and yarn for that specific order.
“For a bulk order where margins are low but consistency is key, AI can help you decide whether you need a slightly different needle or operator setup. All these judgments – the human still has to make them. But AI can tell you what the data says.“
Aminul Islam sees a broader integration coming not just quality inspection, but machine health, needle health, RPM analysis, and even cross-brand benchmarking of knitting machine performance. Dyeing and finishing could follow.
He also points to a trend already in motion: knitting machine manufacturers are beginning to embed AI capabilities into their equipment. “When we recently bought new machines, the manufacturer was already trying to integrate AI technology. In the future, when you buy a knitting machine, the AI may come built in not as a separate system from a separate company.“
The discussion also highlighted that successful implementation depends on continuous learning, factory-specific training data, and close collaboration between technology developers and industry experts. As data volumes increase and systems become more refined, AI is expected to move rapidly from standalone inspection tools to fully integrated components of knitting, dyeing, and finishing operations.
ReCIRCLE by TexSPACE is an initiative to promote dialogue, collaboration, and innovation toward a circular textile ecosystem. Each episode features thought leaders, technology pioneers, brand representatives, and industry practitioners sharing their insights and best practices in textile circularity driving the industry toward a more intelligent, efficient, and sustainable future.
-
-
-
-
