產品與服務
我們提供高效能和可靠性的自動化產品和解決方案
應用案例
我們提供高效能和可靠性的自動化產品
和解決方案
最新消息
偲倢不僅僅是軟體開發公司,我們也致力於自動化
產業的發展,與自動化相關的應用夥伴並肩合作。
07
2021年6月
Advantech MIC-730AI Review NVIDIA Jetson Xavier AI Edge Appliance
Advantech MIC 730AI Passive Heatsink There are few reviews where we genuinely go into so much effort to test a product. In our Advantech MIC-730AI review, we are going to great lengths to show off the innovative solution that the company has. Built around the NVIDIA Jetson Xavier platform, we transformed the STH studio into a textile defect inspection plant. In our review, we are going to show how this NVIDIA AI edge inferencing machine works, and then we are going to show it being used. Advantech MIC-730AI Hardware Overview The front of the Advantech unit contains all of the I/O, and that I/O is a big factor in these types of edge systems. Moving from the left to right we have the power button and an expansion slot on the bottom. We also have a 16-bit DI/DO block in green. Next to those are two COM ports. The two LAN ports are both Gigabit. One is a Marvell 88E1512 and the other is an Intel i210. Advantech MIC 730AI IO There are two USB 2 and two USB 3 ports. One will see on either side there are holes to help with custom connectors retain the USB Type-A plugs. One can also see a HDMI video output and the power delivery/ grounding points. The entire unit is 192mm x 230mm x 87mm and a big portion of that is the other major external chassis feature: the passive cooling heatsink. This passive heatsink is designed to allow the system to run without fans in up to 55C-60C operating temperatures depending on the operating modes. Advantech MIC 730AI Passive Heatsink That heatsink is largely cooling the NVIDIA Jetson Xavier SOM. This SOM sits atop the main system PCB that also has features such as the ports shown above. Advantech MIC 730AI NVIDIA Jetson Xavier The NVIDIA Jetson Xavier SoC is designed to provide AI vision acceleration in a single chip package with eight Arm v8.2 Carmel cores and the video encode/ decode acceleration along with a Volta-generation GPU. NVIDIA Jetson AGX Xavier SoC The CPU also gets 32GB of LPDDR4X memory on the module. In some ways, this is not too dissimilar to NVIDIA’s version of the Apple M1 chip, except that NVIDIA is focusing on video AI and Apple the desktop market. NVIDIA’s CPU cores are not overly fast with integer performance around that of an Intel Xeon E3-1240. On the other hand, the Volta generation GPU means we get tensor cores onboard. This is not meant to be a desktop replacement CPU, rather the CPU is designed more to enable the use of the GPU and accelerators. Advantech MIC 730AI NVIDIA Jetson Lscpu Ubuntu 18.04 LTS NVIDIA Desktop 800 Here is the quick layout of the NVIDIA Jetson Xavier CPU layout. One can see 2MB of L2 cache per two NVIDIA Arm Carmel cores and then a 4MB L3 shared cache. NVIDIA Jetson Xavier CPU Complex On the bottom side of the motherboard has a lot of internal I/O expansion. We normally start from the top and move to the bottom, but here instead we are going to move from the bottom up. Advantech MIC 730AI Internal Overview At the bottom edge in our orientation, we have the DC power board. One likely saw the DC power input on the port overview section. This is the internal side of that connector. We will quickly note for some of our readers less versed in the embedded space that it is possible to have an AC adapter for desktop use even with the industrial versus barrel-type power input connector. The industrial version is simply designed to be more secure and prevent accidental removal. Advantech MIC 730AI Power Board Above this power board, we get the connection to the main PCB. There are PCIe x8 and x16 expansion slots above this but enabling their use requires getting into options that we did not have for testing. Advantech MIC 730AI Expansion Slot There is a small feature we wanted to point out on the left side above. Below the internal reset button, there is an OTG MicroUSB port that one may miss if just casually looking at the photo. On the right side, there are SATA data and power headers. Advantech MIC 730AI SATA Power And Data Headers One can add a 2.5″ disk such as a SSD by adding cables to these headers. The system itself only has 32GB of eMMC storage standard, so this can be a big deal. Advantech MIC 730AI SATA Power And Data Cables On the top edge of the motherboard in our orientation, we get a SD card slot on the left side. There is a MiniPCIe slot in white. Then we get a M.2 2280 (80mm) slot for NVMe storage. Advantech MIC 730AI MiniPCIe And M.2 With SD From an operating perspective, it is important that we do not have any fans here. For industrial applications passive cooling means a lower chance of failure as fans, especially when ingesting debris, can be a point of failure. We also want to quickly note that this unit has been shipped around the world a few times. One can see the wear on some of the metal panels such as in the corners of the photo below. Despite this use, the system still works perfectly fine. Still, we wanted to try to recreate a use case where we may find a system like this. Next, we are going to turn the STH studio into a textile factory. Advantech MIC-730AI Textile / OCR Demo with Smasoft When we decided to do this review, one of the big questions I had was the “so what?” We normally have a performance section, but with CPU performance more akin to a Xeon E3-1240 that seemed less exciting. A 30W passively cooled system is never going to reach the levels of raw performance of 280W CPUs and 500W+ GPUs that we have today. So as a result, we managed to turn the STH studio into a textile inspection factory. Advantech MIC 730AI Smasoft Demo Here is the overall setup. Just a quick thank you to the Advantech and Smasoft folks for helping with this. I felt that I wanted to show off what a solution looks like and they certainly helped make it happen. Advantech MIC 7700 MIC 730AI You will quickly notice we have not one, but two Advantech units. Alongside this MIC-730AI NVIDIA Jetson-based unit, we had an Advantech MIC-7700 Intel-based solution running Windows and the Smasoft software. We will just note that the MIC-7700 was quite loud and used fans, so these are different classes of embedded systems. The basic idea is that the Advantech MIC-730AI to x86 box does not need to be a 1:1 ratio. Instead of adding expensive and power-hungry GPUs to every system, one can use and scale with the Jetson-based solutions. Advantech MIC 730AI Camera Next, we have the camera. The camera is a PoE industrial camera that is held in a fixed position on a stand. There are a number of ways to get power, and Advantech has PoE solutions as well, but here we have a simple PoE injector. Advantech MIC 730AI Camera PoE Injector The Smasoft software also used the camera. This would normally be mounted on a production line, but in our case, the IP camera is being mounted on a test rig so we can photograph it. Advantech Smasoft Robot Camera Below the camera, we have a ring light to get enough light for the camera. Advantech MIC 730AI Ring Light Below this, we have space where we are running our “textile line” and also have a sample OCR page that is being lit by the light and captured via the camera. The Smasoft solution then uses the setup to identify characters. Advantech MIC 730AI Smasoft OCR Fabric can be passed under the camera. One can imagine this is being moved past the camera on a long line, but for the demo, we only have a small piece here. Advantech MIC 730AI Smasoft Fabric Inspection The Smasoft solution can identify a defect in the fabric, or a specific set of text from the OCR demo, and then perform actions on the line to tag what is going across. Fabric Defect Highlighted An example of how this tagging is used is to trigger automation tools. Perhaps to pick the defective part off of a line. One item we did not have was a robot arm. Luckily, Smasoft had an emulator for the robot arm that may be doing picks based on the OCR/ image recognition. Advantec Smasoft Robot Arm Overall, this is a fairly cool solution. Hopefully, this helps you visualize what a more end-to-end solution looks like versus just looking at a box. We know a lot of our readers want to see these systems in action. Unfortunately, we do not have a full factory, but we can set up a small textile mill. As one can imagine, the Smasoft folks have solutions for different industries that are based on the Advantech compute platforms like the MIC-730AI we see here. Advantech MIC-730AI Power Consumption and Noise Just as a quick note here, on 120V AC power we saw up to 30W of power consumption in our demo. NVIDIA’s Jetson line has configurable TDP so one can lower the power consumption at the expense of some performance. Although 30W was the maximum, we saw idle closer to 15W. Advantech MIC 730AI FSP AC DC PSU On the noise front, this unit was very quiet. We had a passively cooled chassis with a passively cooled FSP external AC adapter. In a 24dba ambient environment, we could not register even a 1dba increase from coil whine on our trusty Extech sound meter so we are going to call this silent for our purposes. Final Words Overall, this was a fun solution to take a look at. We wanted to give a better sense of not just the physical solution, but also how it is used in a broader setup. Indeed, the NVIDIA Jetson Xavier platform is popular, and going beyond the developer kit to something that can be deployed. Advantech MIC 730AI Passive Heatsink We are at a time when there is an intense focus on the edge and specifically taking the AI models that have been trained in data centers and putting them to use. AI has gone from a curiosity to being deployed every day. That is really the market for the MIC-730AI. Advantech MIC 730AI NVIDIA Jetson Lscpu Desktop 800 Overall, this is a very cool solution. Having a fanless embedded platform that is designed for a wide range of environments means that one can bring GPU accelerated computing to the edge. Source: STH
07
2021年6月
AI成製造業搶單關鍵 解決方案商如何出奇致勝?
▲AI運用於檢測設備的解決方案相繼問世,吸引業者躍躍欲試。偲倢科技 以機器取代人力的視覺檢測技術,堪稱製造業運用AI技術最廣泛的領域,尤其AI運用於檢測設備的解決方案相繼問世後,無論是購置成本或檢測效能,都吸引業者躍躍欲試,甚至是否導入AI技術,更成為業者未來能否拿下訂單的關鍵之一,而面對市場快速增長的需求,也刺激解決方案市場催生更多元的服務模式。 當前製造業者對於引進自動化的智能檢測系統有高度興趣,但即便如此,業者往往仍會遇到幾個問題,專注於工業自動化的軟體開發的AI新創偲倢科技長期與台灣中小企業交手,而從執行長陳青煒的經驗中觀察就發現,這類型企業大部分因為專業人才團隊建立不易,且又要跟既有團隊融合,以及導入周期過長等因素,面對無法跨越的門檻,這也是為什麼企業需要引進外援的主要原因。 目前在智慧製造的市場上,視覺檢測是應用最廣泛的領域,但隨著製造業者躍躍欲試,越來越多問題也一一浮現,對於解決方案供應商來說,如何達到客戶預期也頗具挑戰。 舉例來說,都知道AI是由資料驅動,因此如何取得資料是導入AI的先決條件。以AI品質檢測來說,像是產品的外觀成像資料就是餵養AI的大數據,因此陳青煒表示:「有的客戶過去沒有蒐集這類資料的習慣,因此在正式進入影像分析前,取像環境的建置同時也是供應商必須具備的能力之一,而這也是業者往往必須花更多心力投入的過程。」 事實上在經過取像與蒐集資料之後,業者的挑戰才正要開始。陳青煒說:「未來業者還要面臨如何讓AI落地、落地後怎麼大規模複製到其他場域佈署,以及如何管理這些大量的AI模型,甚至在跨廠區之間管理不同的設備資料等。」而這些挑戰側面來說,也刺激著解決方案供應商催生更多元的服務模式。 像是供應商勢必要走向以模組化的形式來迎合不同產業類別的客戶需求,如何快速組成符合應用情境的解決方案,成為供應商立足市場的必要之一;其次,則是透過軟體加值的方式,為既有的AOI設備賦能,讓製造者不需要修改現有產線設計,就能享受AI加值的效益。 隨著自動化需求漸盛,AI成為製造業者提升競爭力的關鍵。有的業者藉此來填補缺口越來越龐大的人力需求,但陳青煒坦言:「以AI質檢應用來說,省多少人力並不一定是重點,目前既有的AOI設備有的也已經完全不需要人工複檢,但AI可以解決AOI過往過殺率(Overkill)過高的問題,不管是過殺造成的重工或是被白白丟棄的良品,事實上都是一種無形的浪費,如果跳脫ROI的框架來看,這往往比節省人力更具有一定的價值。」 除此之外,從市場端反饋的訊息來看,越來越多終端客戶也開始將供應鏈是否導入AI作為訂單的附帶條件,陳青煒觀察,這似乎已成為一種未來的主流,有的終端客戶不見得懂AI,但不得不承認的是,AI確實已在全球各行各業掀起一股潮流,除了提升產品品質,現在AI更成為一種正面宣傳的手法之一。 Source: DIGITIMES
02
2021年6月
Computex 2021:NVIDIA 邀請台灣夥伴矽品、英業達、偲倢科技與小柿智檢分享智慧製造應用
NVIDIA 除了 Computex 的主題演講外,也參與多場分場專題,而 NVIDIA 亦針對其中幾個主題邀請台灣的合作夥伴探討如何透過 NVIDIA 的技術創造產業升級。稍早  NVIDIA 針對智慧製造領域,邀請矽品、英業達兩家自建智慧製造應用的大型企業,與偲倢科技、小柿智檢兩家協助企業導入 AI 智慧製造解決方案的公司進行分享。 四家台灣廠商對於 AI 在智慧製造的觀點相當類似,皆認為以 AI 取代人力並非 AI 真正的目的,而是使 AI 能減少以人力進行檢測時產生的人力疲勞、或是倦怠後產生評估準確率下降的問題,同時利用 AI 技術評估部分瑕疵的可用性,提升產品的產能與良率;四家廠商也分別扮演具備規模的大企業如何建構自身的 AI 技術團隊,以及具備 AI 與系統整合技術的服務商如何協助中小企業實現智慧製造。 矽品 ▲矽品目前面臨的問題是模型的進步幅度比預期慢,還有後續 AI 的可靠度管理問題 矽品作為封裝測試大廠,早在先前就開始評估導入 AI 技術作為封裝的機器視覺檢測,當前使用 NVIDIA 的 DGX 2 作為 AI 模型訓練,並以基於 8 張 NVIDIA T4 的 AI 系統作為場域佈署的基礎建設;當前矽品針對不同的產品類型約訓練達 200 種的 AI 模型,面臨的挑戰在兩部分,雖然當前模型辨識的精確度不錯,但模型訓練的提升幅度比預期的慢,正確率在五成到九成不等,其次是檢測系統的管理與維護,畢竟當前矽品對於 AI 的發展仍未達完全信任,在模型導入測試一段時間後仍需重複驗證 AI 模型的準確率;矽品也不諱言投入 AI 的前一年半幾乎沒有突破,但後續在 NVIDIA 持續協助下的發展速度終於獲得起飛,雖然目前 AI 技術的導入比起預期來的慢,但矽品仍看好 AI 模型後續帶來的價值。 英業達 ▲英業達主要應用於產品生產的 QC 品管檢測 ▲英業達強調自研模型具備自適應能力,利用單一模型可以適應不同的產品 英業達主要將 AI 應用在產品的製造檢測領域,諸如筆電產品、 PCB 的外觀檢測,同時也建構全台第一座可技術輸出的 5G ORAN 智慧工廠。英業達開發一套用於產品外觀檢測的通用模型,能夠自適應不同的產品設計進行判斷,關鍵在於英業達的 AI 模型不同於傳統外觀判斷的 AI 模型訓練是標註瑕疵類型,而是使 AI 能夠判斷完好生產的商品該有的外觀,其 AI 模型僅需判斷外觀是否符合預期,檢測外觀是否有瑕疵問題,簡化判斷程序。此外英業達的 AI 團隊也將 AI 用於研判業界趨勢與原物料供應,藉 AI 分析預估市場需求變化、零組件波動,率先進行調整,使生產更為順遂。 偲倢科技 SmaSOFT ▲偲倢的解決方案以四大模組構成,因應規模與需求調整與擴充 ▲藉由客戶類型的不同彈性佈署軟體模組 ▲半導體平整度檢測的客戶案例 ▲金屬加工件的良率判斷 偲倢科技是一家提供 AI 與 AOI 軟體解決方案的軟體自動化公司,主要以協助不具 AI 規劃能力的中小規模企業導入 AI 技術,並幫助他們自訓練資料、 AI 落地到管理;偲倢科技原先是以機器手臂與視覺自動化方案為主,後續因應客戶需求提供 SmaMAX 的跨廠區資料管理,以及 SmaONE 雲平台資料管理,並透過旗下四大軟體模組的組合,因應客戶規模、應用類型進行組合。此次偲倢也舉出旗下兩個合作客戶案例說明智慧製造帶來的價值,其一是提供半導體客戶進行表面平整度的檢測,透過兩種 AI 模型分辨六種瑕疵,除了有 99.5% 的檢測準確率以外,也進一步從檢測結果使生產良率自 85% 提高度 93% 。另一個則是傳統加工製造的車用金屬加工件案例,透過 AOI 檢測轉化為 AI 檢測,除了檢查金屬加工品的品質以外,更透過 AI 判斷瑕疵是否影響功能性,進一步減少以往透過人工檢測過於嚴格的過殺率,也降低大量的人檢時間。 ▲NVIDIA Executive Keynote (SmaSOFT)—41:39s 小柿智檢 ▲小柿智檢的訓練平台可透過雲進行,並以基於 NVIDIA T4 的系統進行佈署 ▲相較傳統的影像檢測, AI 檢測能判讀更多的生產瑕疵 ▲IPC 公司透過小柿智檢的落地應用專用機檢查機殼的外觀瑕疵 ▲針對小型金屬加工物的 AI 六面檢測 小柿智檢是由富士康集團內部創新獨立的新創公司,是針對中小企業建構以軟體為核心的 AI 檢測方案,定位在一半軟體、一半設備商,並藉由 AWS 雲平台能夠減少企業佈署 AI 訓練系統的負擔,但企業亦可選擇自行擁有 AI 訓練平台,另外搭配基於 NVIDIA T4 的終端設備實現 AI 落地,同時小柿智檢可提供多元的感測器類型,符合不同產業影像辨識與檢測的需求。小柿智檢也舉了幾項應用案例,例如中部製鞋廠透過 AI 影像檢測,除了能檢驗傳統影像辨識技術也能判斷的的色差與差排缺件問題外,進一步可找出如包風、短射、刮傷、毛邊、氣泡等生產瑕疵;另外在 IPC 廠則透過落地應用專機可檢查工業電腦的外觀生產瑕疵,並透過機器手臂的彈性打光優點,能夠有更高的判斷精確度;最後則是透過落地應用專機進行小尺寸加工件的 AI 六面檢測,可判斷如裂痕、髒汙、崩角與崩邊等情況。 Source: COOL3C            @NVIDIA YouTube Channel